Does your business struggle to understand the data in a better way or to predict future trends? Then you’re not the only one in the business; many fail here. ARIMA can help you forecast and understand the new patterns from the past data using time series analysis. One of the top reasons why the ARIMA model is always in demand is that lagged moving averages smooth the time series data.
You mostly witness this method in technical analysis to forecast future security prices. To get a better idea how this works, you need to understand several core topics:
Time Series Forecasting
Time series forecasting is a trend analysis technique that focuses on cyclical fluctuations analysis, and seasonality issues go through the past data and associated patterns to predict the future trend. Success is not guaranteed in this method, though it throws a hint about future trends.
Time series forecasting uses Box-Jenkins Model, which involves three methods to predict future data: autoregression, differencing and moving averages (also called p, d, q, respectively).
The Box-Jenkins model is an advanced technique to forecast based on input data from the specified time series and conjointly conferred as an autoregressive integrated moving average method ARIMA (p,d,q). Using the ARIMA model, you can forecast a time series using the past series values.
The best uses of ARIMA models are to forecast stock prices and earnings growth.
Nomenclature in ARIMA Model
As an ARIMA(p,d,q) model, a nonseasonal ARIMA is one that:
prepresents the number of autoregressive terms,
dis the necessary number of nonseasonal changes for stationarity
qis the number of lags in the prediction equation.
In terms of y, the general forecasting equation is:
the first-order autoregressive model, if the series is stationary and autocorrelated, it’s predicted as the simple multiple of its previous value and a constant. And the equation becomes:
Ŷt = μ + ϕ1Yt-1
Y then regressed on itself after lagging by one period, meaning Y = 0, plus a constant term.
If the slope coefficient Ф1 is positive and less than 1 in magnitude, the model shows mean-reverting behavior in which the next predicted value to be Ф1 times as far away from the mean as this period’s value. If Ф1 is negative, the model shows the mean-reverting behavior with alternation of signs. And Y will be below the average next period if it is above the same period.
ARIMA (0,1,1) with constant:
After implementing the SES model as the ARIMA model, it gains flexibility; first, the estimated MA (1) coefficient allowed to be negative: corresponds to a smoothing factor more prominent than 1, which forbids in SES model-fitting procedure. Second, you can add a constant term in the ARIMA model to estimate an average non-zero trend.
Ŷt = μ + Yt-1 –θ1et-1
How to Make a Series Stationary in Time Series Forecasting?
The most simplified approach to make it stationary is to differentiate it and subtract the previous value from the correct value. Depending upon the series complexity, you may require more than one differentiation.
The value of d has to be the minimum number of differentiating to make the series stationary. Therefore the value of d has to be zero, i.e. (d = 0)
AR and MA Models in terms of (p), (q), and (d):
AR(p): AutoRegression: a robust model that uses the dependent relationship between the current observation and previous observations. It utilizes the past values in the regression equation for thetime series forecasting.
I (d) Integration: Makes the process stationary with a differentiation (subtracting the previous value from the current value for the d number of times till it becomes (d = 0)
MA (q): Moving Average: utilize the dependency between an observation and a residual error from the moving average model when applied to lagged observation. The moving average method draws the error of the model as the combination of previous faults. And the order q represents the number of terms in the model.
How to Handle if a Time Series Analysis is Slightly Under or Over Differenced:
The time series method at this point may be slightly under-differentiated, and when you differentiate it one more time, it can then become over-differentiated. When the series is under-differentiated, adding one or more additional AR terms usually makes it up. And when it is over-differentiated, try adding further MA terms to get the balance.
Accuracy Metrics in Time Series Analysis
The commonly used accuracy metrics to evaluate the accuracy of the forecasting:
Mean Absolute Percentage Error (MAPE)
Mean Error (ME)
Mean Absolute Error (MAE)
Mean percentage Error (MPE)
Root Mean Square Error (RMSE)
Lag 1 Autocorrelation of Error (ACF1)
Correlation between the Actual and the Forecast (Corr)
Min-Max Error (MinMax)
Final Words
Time series forecasting is a classic method to understand futuristic trends and patterns, although success is not guaranteed. But to be at the top, businesses need regular analysis of previous and ongoing trends to understand the future trends, and that’s where time series forecasting comes into action.
And Time Series Forecasting ARIMA model uses autoregression and moving averages methods to predict the accurate results followed by accuracy metrics. In a nutshell, you learned in-depth about the ARIMA model, its terminology, making a series stationary, handling time series under and over differentiated, followed by the accuracy metrics.
Does your business struggle to understand the data in a better way or to predict future trends? Then you’re not the only one in the business; many fail here. ARIMA can help you forecast and understand the new patterns from the past data using time series analysis. One of the top reasons why the ARIMA model is always in demand is that lagged moving averages smooth the time series data.
You mostly witness this method in technical analysis to forecast future security prices. To get a better idea how this works, you need to understand several core topics:
Time Series Forecasting
Time series forecasting is a trend analysis technique that focuses on cyclical fluctuations analysis, and seasonality issues go through the past data and associated patterns to predict the future trend. Success is not guaranteed in this method, though it throws a hint about future trends.
Time series forecasting uses Box-Jenkins Model, which involves three methods to predict future data: autoregression, differencing and moving averages (also called p, d, q, respectively).
The Box-Jenkins model is an advanced technique to forecast based on input data from the specified time series and conjointly conferred as an autoregressive integrated moving average method ARIMA (p,d,q). Using the ARIMA model, you can forecast a time series using the past series values.
The best uses of ARIMA models are to forecast stock prices and earnings growth.
Nomenclature in ARIMA Model
As an ARIMA(p,d,q) model, a nonseasonal ARIMA is one that:
prepresents the number of autoregressive terms,
dis the necessary number of nonseasonal changes for stationarity
qis the number of lags in the prediction equation.
In terms of y, the general forecasting equation is:
the first-order autoregressive model, if the series is stationary and autocorrelated, it’s predicted as the simple multiple of its previous value and a constant. And the equation becomes:
Ŷt = μ + ϕ1Yt-1
Y then regressed on itself after lagging by one period, meaning Y = 0, plus a constant term.
If the slope coefficient Ф1 is positive and less than 1 in magnitude, the model shows mean-reverting behavior in which the next predicted value to be Ф1 times as far away from the mean as this period’s value. If Ф1 is negative, the model shows the mean-reverting behavior with alternation of signs. And Y will be below the average next period if it is above the same period.
ARIMA (0,1,1) with constant:
After implementing the SES model as the ARIMA model, it gains flexibility; first, the estimated MA (1) coefficient allowed to be negative: corresponds to a smoothing factor more prominent than 1, which forbids in SES model-fitting procedure. Second, you can add a constant term in the ARIMA model to estimate an average non-zero trend.
Ŷt = μ + Yt-1 –θ1et-1
How to Make a Series Stationary in Time Series Forecasting?
The most simplified approach to make it stationary is to differentiate it and subtract the previous value from the correct value. Depending upon the series complexity, you may require more than one differentiation.
The value of d has to be the minimum number of differentiating to make the series stationary. Therefore the value of d has to be zero, i.e. (d = 0)
AR and MA Models in terms of (p), (q), and (d):
AR(p): AutoRegression: a robust model that uses the dependent relationship between the current observation and previous observations. It utilizes the past values in the regression equation for thetime series forecasting.
I (d) Integration: Makes the process stationary with a differentiation (subtracting the previous value from the current value for the d number of times till it becomes (d = 0)
MA (q): Moving Average: utilize the dependency between an observation and a residual error from the moving average model when applied to lagged observation. The moving average method draws the error of the model as the combination of previous faults. And the order q represents the number of terms in the model.
How to Handle if a Time Series Analysis is Slightly Under or Over Differenced:
The time series method at this point may be slightly under-differentiated, and when you differentiate it one more time, it can then become over-differentiated. When the series is under-differentiated, adding one or more additional AR terms usually makes it up. And when it is over-differentiated, try adding further MA terms to get the balance.
Accuracy Metrics in Time Series Analysis
The commonly used accuracy metrics to evaluate the accuracy of the forecasting:
Mean Absolute Percentage Error (MAPE)
Mean Error (ME)
Mean Absolute Error (MAE)
Mean percentage Error (MPE)
Root Mean Square Error (RMSE)
Lag 1 Autocorrelation of Error (ACF1)
Correlation between the Actual and the Forecast (Corr)
Min-Max Error (MinMax)
Final Words
Time series forecasting is a classic method to understand futuristic trends and patterns, although success is not guaranteed. But to be at the top, businesses need regular analysis of previous and ongoing trends to understand the future trends, and that’s where time series forecasting comes into action.
And Time Series Forecasting ARIMA model uses autoregression and moving averages methods to predict the accurate results followed by accuracy metrics. In a nutshell, you learned in-depth about the ARIMA model, its terminology, making a series stationary, handling time series under and over differentiated, followed by the accuracy metrics.
If you are wondering how much a data scientist earns, whether you are a hiring manager or looking for a job, there are plenty of websites providing rather detailed information, broken down by area, seniority, and skills. Here I focus on the United States, offering a summary based on various trusted websites.
A starting point is LinkedIn. Sometimes, the salary attached to a position is listed, and LinkedIn will tell you how many people viewed the job ad, and how well you fit based on skill matching and experience. LinkedIn will even tell you which of your connections work for the company in question, so you may contact the most relevant ones. Positions with fewer views, that are two week old, are less competitive (but maybe less attractive too), but if you don’t have much experience, they could be worth applying to. You probably receive such job ads in your mailbox, from LinkedIn, every week. If not, you need to work on your LinkedIn profile (or maybe you don’t want to receive such emails).
Popular websites with detailed information include PayScale, GlassDoor, and Indeed. GlassDoor, based on 17,000 reported salaries (see here), mentions a range from $82k to $165k, with an average of $116k per year for a level-2 data scientist. It climbs to $140k for level-3. You can do a search by city or company. Some companies listed include:
Facebook: $153,000 based on 1,006 salaries. The range is $55K – $226K.
Quora: $122,875 based on 509 salaries. The range is $113K – $164K.
Oracle: $148,396 based on 457 salaries. The range $88K – $178K.
IBM: $130,546 based on 382 salaries. The range is $58K – $244K.
Google: $148,560 based on 246 salaries. The range is $23K – $260K.
Microsoft: $134,042 based on 204 salaries. The range is $13K – $292K.
Amazon: $125,704 based on 190 salaries. The range is $60K – $235K.
Booz Allen Hamilton: $90,000 based on 186 salaries. The range is $66K – $215K.
Walmart: $108,937 based on 185 salaries. The range is $78K – $186K.
Cisco: $157,228 based on 166 salaries. The range is $79K – $186K.
Uber: $143,661 based on 137 salaries. The range is $56K – $200K.
Intel: $125,936 based on 129 salaries. The range is $58K – $180K.
Apple: $153,885 based on 128 salaries. The range is $60K – $210K.
Airbnb: $180,569 based on 122 salaries. The range is $99K – $242K.
These are base salaries and do not include bonus, stock options, or other perks. Companies with many employees in the Bay Area offer bigger salaries due to the cost of living. These statistics may be somewhat biased as very senior employees are less likely to provide their salary information. A chief data scientist typically makes well above $200k a year, not including bonuses, and an $800k salary, at that level, at companies such as Microsoft or Deloitte (based on my experience), is not uncommon. On the low end, you have interns and part-time workers. If you visit Glassdoor, you can get much more granular data.
Below are statistics coming this time from Indeed (see here). They offer a different perspective, with breakdown by type of expertise and area. The top 5 cities with highest salaries are San Francisco ($157,041), Santa Clara ($156,284), New York ($140,262), Austin ($133,562) and San Diego ($124,679). Surprisingly, the pay is lower in Seattle than in Houston. Note that if you work remotely for a company in the Bay Area, you may get a lower salary if you live in an area with lower cost of living. Still, you would be financially better off than your peers in San Francisco.
The kind of experience commanding the highest salary (20 to 40% above average) are Cloud Architecture, DevOps, CI/CD (continuous delivery and/or continuous deployment), Microservices, and Performance Marketing. Finally, Indeed also displays salaries for related occupations, with the following averages:
The average for Data Scientist is $119,444 according to Indeed. This number is similar to the one coming from Glassdoor. Note that some well-funded startups can offer large salaries. My highest salary was as chief scientist / co-founder at a company with less than 20 employees. And my highest compensation was for a company I created and funded myself, though I was not on a payroll and I did not assign myself a job title.
To receive a weekly digest of our new articles, subscribe to our newsletter, here.
About the author: Vincent Granville is a data science pioneer, mathematician, book author (Wiley), patent owner, former post-doc at Cambridge University, former VC-funded executive, with 20+ years of corporate experience including CNET, NBC, Visa, Wells Fargo, Microsoft, eBay. Vincent is also self-publisher at DataShaping.com, and founded and co-founded a few start-ups, including one with a successful exit (Data Science Central acquired by Tech Target). He recently opened Paris Restaurant, in Anacortes. You can access Vincent’s articles and books, here.
Young and dynamic data science and machine learning enthusiasts are all are very interested in making a career transition by learning and doing as much hands-on learning as possible with these technologies and concepts as Data Scientist or Machine Learning Engineers or Data Engineers or Data Analytics Engineers. I believe they must have the Project Experience and a job-winning portfolio in hand before they hit the interview process.
Certainly, this interview process would be challenging, NOT only for the freshers, but also for experienced individuals since these are all new techniques, domain, process approach, and implementation methodologies that are totally different from traditional software development. Of course, we could adopt an agile mode of delivery and no excuse from modern cloud adoption techniques and state beyond all industries and domains, who are all looking and interested in artificial intelligence and machine learning (AI and ML) and its potential benefits.
In this article, I will to discuss how to choose the best data science and ML projects during the capstone stages of your schools, colleges, training institutions, and specific job-hunting perspective. You could map this effort with our journey towards getting your dream job in the data science and machine learning industry.
Without further ado, here are the top 20 machine learning project that can help you get started in your career as a machine learning engineer or data scientist that can be a great add-on to your portfolio.
1. Data Science Project – Ultrasound Nerve Segmentation
Problem Statement & Solution
In this project, you will be working on building a machine learning model that can identify nerve structures in a data set of ultrasound images of the neck. This will help enhance catheter placement and contribute to a more pain-free future.
Even the bravest patients cringe at the mention of a surgical procedure. Surgery inevitably brings discomfort, and oftentimes involves significant post-surgical pain. Currently, patient pain is frequently managed using narcotics that bring a number of unwanted side effects.
This data science project’s sponsor is working to improve the pain management system using indwelling catheters that block or mitigate pain at the source. These pain management catheters reduce dependence on narcotics and speed up patient recovery.
The project objective is to precisely identify the nerve structures in the given ultrasound images, and this is a critical step in effectively inserting a patient’s pain management catheter. This project has been developed in python language, so it is easy to understand the flow of the project and the objectives. They must build a model that can identify nerve structures in a dataset of given ultrasound images of the neck. Doing so would improve catheter placement and contribute to a more pain-free future.
Let see the simple workflow.
Certainly, this project would help us to understand the image classification and highly sensitive area of analysis in the medical domain.
Take away and outcome and of this project experience.
Understanding what image segmentation is.
Understanding of subjective segmentation and objective segmentation
The idea of converting images into matrix format.
How to calculate euclidean distance.
Scope of what dendrogram are and what they represent.
Overview of agglomerative clustering and its significance
Knowledge of VQmeans clustering
Experiencing grayscale conversion and reading image files.
A practical way of converting masked images into suitable colours.
How to extract the features from the images.
Recursively splitting a tile of an image into different quadrants.
2. Machine Learning project for Retail Price Optimization
Problem Statement
In this machine learning pricing project, we must implement retail price optimization and apply a regression trees algorithm. This is one of the best ways to build a dynamic pricing model, so developers can understand how to build models dynamically with commercial data which is available from a nearby source and visualization of the solution is tangible.
Solution Approach: In this competitive business world “PRICING A PRODUCT” is a crucial aspect. So, we must gather a lot of thought process into that solution approach. There are different strategies to optimize the pricing of products. And must take extra care during the pricing of the products due to their sensitive impact on the sales and forecast. While there are products whose sales are not very affected by their price changes, they could be luxury items or essentials products in the market. This machine learning retail price optimization project will focus on the former type of products.
This project clearly captures the data and aligns with the “Price Elasticity of Demand” phenomenon. This exposes the degree to which the effective desire for something changes as its price the customers desire could drop sharply even with a little price increase, I mean directly proportional relationship. Generally, economists use the term elasticity to denote this sensitivity to price increases.
In this Machine Learning Pricing Optimization project, we will take the data from the café shop and, based on their past sales, identify the optimal prices for their list of items, based on the price elasticity model of the items. For each café item, the “Price Elasticity” will be calculated from the available data and then the optimal price will be calculated. A similar kind of work can be extended to price any products in the market.
Take away and Outcome and of this project experience.
Understanding the retail price optimization problem
Understanding of price elasticity (Price Elasticity of Demand)
Understanding the data and feature correlations with the help of visualizations
Understanding real-time business context with EDA (Exploratory Data Analysis) process
How to segregate data based on analysis.
Coding techniques to identify price elasticity of items on the shelf and price optimization.
3. Demand prediction of driver availability using multistep Time Series Analysis
Problem Statement & Situation:
In this supervised learning machine learning project, you will predict the availability of a driver in a specific areaby using multi-step time series analysis. This project is an interesting one since it is based on a real-time scenario.
We all love to order food online and do not like to experience delivery fee price variation. Delivery charges are always highly dependent on the availability of drivers in your area in and around, so the demand of orders in your area, and distance covered would greatly impact the delivery charges. Due to driver unavailability, there is an impact in delivery pricing increasing and directly this will hit the many customers who have dropped off from ordering or moving into another food delivery provider, so at the end of the day food suppliers (Small/medium scale restaurants) are reducing their online orders.
To handle this situation, we must track the number of hours a particular delivery driver is active online and where he is working and delivering foods, and how many orders in that area, so based on all these factors certainly, we can efficiently allocate a defined number of drivers to a particular area depending on demand as mentioned earlier.
Take away and Outcome and of this project experience.
How to convert a Time Series problem to a Supervised Learning problem.
What exactly is Multi-Step Time Series Forecast analysis?
How does Data Pre-processing function in Time Series analysis?
How to do Exploratory Data Analysis (EDA) on Time-Series?
How to do Feature Engineering in Time Series by breaking Time Features to days of the week, weekend.
Understand the concept of Lead-Lag and Rolling Mean.
Clarity of Auto-Correlation Function (ACF) and Partial Auto-Correlation Function (PACF) in Time Series.
Different strategic approaches to solving Multi-Step Time Series problem
Solving Time-Series with a Regressor Model
How to implement Online Hours Prediction with Ensemble Models (Random Forest and Xgboost)
4. Customer Market Basket Analysis using Apriori and FP- growth algorithms
Problem Statement & Solution
In this project, anyone can learn how to perform Market Basket Analysis (MBA) with the application of Apriori and FP growth algorithms based on the concept of association rule learning, one of my favorite topics in data science.
Mix and Match is a familiar term in the US, I remember I used to get the toys for my kid. It was the ultimate experience you know. Same time keeping things together nearby, like bread and jam–shaving razor and cream, these are the simple examples for MBA, and this is making the customer buy additional purchases more likely.
It is a widely used technique to identify the best possible mix of products or services that comes together commonly. This is also called “Product Association Analysis” or “Association Rules”. This approach is best fit physical retail stores and even online too. In other ways, it can help in floor planning and placement of products.
Take away and Outcome and of this project experience.
Understanding of Market Basket Analysis and Association rules
For the Apriori algorithm & FP- growth algorithm
Exploratory Data Analysis – Univariate & Bivariate analysis
Creating baskets for analysis
Gaining the knowledge on Apriori and FP- growth algorithm
5. E-commerce product reviews – Pairwise ranking and sentiment analysis.
Problem Statement & Solution
Product recommendation systems for the products which are sold over the online-based pairwise ranking and sentiment analysis. So, we are going to perform sentiment analysis on product reviews given by the customers who are all purchased the items and ranking them based on weightage. Here, the reviews play a vital role in product recommendation systems.
Obviously, reviews from customers are very useful and impactful for customers who are going to buy the products. Generally, a huge number of reviews in the bucket would create unnecessary confusion in the selection and buying interest on a specific product. If we have appropriate filters from the collective informative reviews. This proportional issue has been attempted and addressed in this project solution.
This recommendation work has been done in four phases.
Data pre-processing/filtering
Which includes.
Language Detection
Gibberish Detection
Profanity Detection
Feature extraction,
Pairwise Review Ranking,
The outcome of the model will be a collection of the reviews for a particular product and its ranking based on relevance using a pairwise ranking approach method/model.
Take away and Outcome and of this project experience.
EDA Process
Over Textual Data
Extracted Featured with Target Class
Using Featuring Engineering and extracting relevance from data
Reviews Text Data Pre-processing in terms of
Language Detection
Gibberish Detection
Profanity Detection, and Spelling Correction
Understand how to find gibberish by Markov Chain Concept
Hands-On experience on Sentiment Analysis
Finding Polarity and Subjectivity from Reviews
Learning How to Rank – Like Pairwise Ranking
How to convert Ranking into Classification Problem
Pairwise Ranking reviews with Random Forest Classifier
Understand the Evaluation Metrics concepts
Classification Accuracy and Ranking Accuracy
6. Customer Churn Prediction Analysis using Ensemble Techniques
Problem Statement & Solution
In some situations, the customers are closing their accounts or switching to other competitor banks for to many reasons. This could cause a huge dip in their quarterly revenues and might significantly affect annual revenues for the enduring financial year, this would directly cause the stocks to plunge and the market cap to reduce considerably. Here, the idea is to be able to predict which customers are going to churn, and how to retain them, with necessary actions/steps/interventions by the bank proactively.
In this project, we must implement a churn prediction model using ensemble techniques.
Here we are collecting customer data about his/her past transactions details with the bank and statistical characteristics information for deep analysis of the customers. With help of these data points, we could establish relations and associations between data features and customer’s tendency to possible churn. Based on that, we will build a classification model to predict whether the specific set of customers(s) will indeed leave the bank or not. Clearly draw the insight and identify which factor(s) are accountable for the churn of the customers.
Take away and Outcome and of this project experience.
Defining and deriving the relevant metrics
Exploratory Data Analysis
Univariate, Bivariate analysis,
Outlier treatment
Label Encoder/One Hot Encoder
How to avoid data leakage during the data processing
Understanding Feature transforms, engineering, and selection
Hands-on Tree visualizations and SHAP and Class imbalance techniques
Knowledge in Hyperparameter tuning
Random Search
Grid Search
Assembling multiple models and error analysis.
7. Build a Music Recommendation Algorithm using KKBox’s Dataset.
Problem Statement & SolutionMusic Recommendation Project using Machine Learning to predict the best chances of a user listening and loving a song again after their very first noticeable listening event. As we know, the most popular evergreen entertainment is music, no doubt about that. There might be a mode of listening on different platforms, but ultimately everyone will be listening to music with this well-developed digital world era. Nowadays, the accessibility of music services has been increasing exponentially ranging from classical, jazz, pop etc.,
Due to the increasing number of songs of all genres, it has become very difficult to recommend appropriate songs to music lovers. The question is that the music recommendation system should understand the music lover’s favorites and inclinations to other similar music lovers and offer the songs to them on the go, by reading their pulse.
In the digital market we have excellent music streaming applications available like YouTube, Amazon Music, Spotify etc., All they have their own features to recommend music to music lovers based on their listening history and first and best choice. This plays a vital role in this business to catch the customers on the go. Those recommendations are used to predict and indicate an appropriate list of songs based on the characteristics of the music, which has been heard by music lovers over the period.
This project uses the KKBOX dataset and demonstrates the machine learning techniques that can be applied to recommend songs to music lovers based on their listening patterns which were created from their history.
Take away and Outcome and of this project experience.
Understanding inferences about data and data visualization
Gaining knowledge on Feature Engineering and Outlier treatment
The reason behind Train and Test split for model validation
Best Understanding and Building capabilities on the algorithm below
Logistic Regression model
Decision Tree classifier
Random Forest Classifier
XGBoost model
8.Image Segmentation using Masked R-CNN with TensorFlow
Problem Statement & Solution
Fire is one of the deadliest risk situations. Generally, fire can destroy an area completely in a very short span of time. Another end this leads to an increase in air pollution and directly affects the environment and an increase in global warming. This leads to the loss of expensive property. Hence early fire detection is very important.
The Object of this project is to build a deep neural network model that will give precise accuracy in the detection of fire in the given set of images. In this Deep Learning-based project on Image Segmentation using Python language, we are going to implement the Mask R-CNN model for early fire detection.
In this project, we are going to build early fire detection using the image segmentation technique with the help of the MRCNN model. Here, fire detection by adopting the RGB model (Color: Red, Green, Blue), which is based on chromatic and disorder measurement for extracting fire pixels and smoke pixels from the image. With the help of this model, we can locate the position where the fire is present, and which will help the fire authorities to take appropriate actions to prevent any kind of loss.
Take away and Outcome and of this project experience.
Understanding the concepts
Image detection
Image localization
Image segmentation
Backbone
Role of the backbone (restnet101) in Mask RCNN model
MS COCO
Understanding the concepts
Region Proposal Network (RPN)
ROI Classifier and bounding box Regressor.
Distinguishing between Transfer Learning and Machine Learning.
Demonstrating image annotation using VGG Annotator.
The best understanding of how to create and store the log files per epoch.
9. Loan Eligibility Prediction using Gradient Boosting Classifier
Problem Statement & Solution
In this project, we are predicting if a loan should be given to an applicant or not for the given data of various customers who are all seeking the loan based on several factors like their credit score and history. The ultimate aim is to avoid manual efforts and give approval with the help of a machine learning model, after analyzing the data and processing for machine learning operations. On the top of the machine, the learning solution will look at different factors based on testing the dataset and decide whether to grant a loan or not to the respective individual.
In this ML problem, we use to cleanse the data and fill in the missing values and bringing various factors of the applicant like credit score, history and from those we will try to predict the loan granting by building a classification model and the output will be giving output in the form of probability score along with Loan Granted or Refused as output from the model.
Take away and Outcome and of this project experience.
Understanding in-depth:
Data preparation
Data Cleansing and Preparation
Exploratory Data Analysis
Feature engineering
Cross-Validation
ROC Curve, MCC scorer etc
Data Balancing using SMOTE.
Scheduling ML jobs for automation
How to create custom functions for machine learning models
Defining an approach to solve
ML Classification problems
Gradient Boosting, XGBoost etc
10.Human Activity Recognition Using Multiclass Classification
Problem Statement & Solution
In this project we are going to classify human activity, we use multiclass classification machine learning techniques and analyze the fitness dataset from a smartphone tracker. 30 activities of daily participants have been recorded through a smartphone with embedded inertial sensors and build a strong dataset for activity recognition point of view. Target activities are WALKING, WALKING UPSTAIRS, WALKING DOWNSTAIRS, SITTING, STANDING, LAYING, by capturing 3-axial linear acceleration and 3-axial angular velocity at a constant rate of 50Hz. The objective is to classify activities mentioned above among 6 and 2 different axials. This was captured by an embedded accelerometer and gyroscope in the smartphone. The experiments have been video-recorded to label the data manually. The obtained dataset has been randomly partitioned into two sets as 70% for training and 30% for test data.
Take away and Outcome and of this project experience.
Understanding
Data Science Life Cycle
EDA
Univariate and Bivariate analysis
Data visualizations using various charts.
Cleaning and preparing the data for modelling.
Standard Scaling and normalizing the dataset.
Selecting the best model and making predictions
How to perform PCA to reduce the number of features
Understanding how to apply
Logistic Regression & SVM
Random Forest Regressor, XGBoost and KNN
Deep Neural Networks
Deep knowledge in Hyper Parameter tuning for ANN and SVM.
How to plot the confusion matrix for visualizing the result
Develop the Flask API for the selected model.
Project Idea Credits – ProjectPro helps professionals get their work done faster and with practical experience with verified reusable solution code, real-world project problem statements, and solutions from various industry experts
Machine learning is such a new field that a mature industry-wide standard practice of operations has not yet emerged, like there has been in software development for the past 20 or more years. An ML practitioner who transfers from one company to another would find very big differences in the way each organization brings AI projects to production–if they do at all.
The feature store is an element of data infrastructure that has emerged in the ML community over the past year as a centerpiece of ML pipelines. Adopting a feature store can be a force multiplier for companies trying to transform with data science.
The feature store is not about storing features. A feature store is much more than simply a repository for features, it’s a system that runs scalable, high-performance data pipelines to transform raw data into features. With this system, ML teams can define features once, and deploy to production without rewriting.
And yes, a feature store also:
Catalogs and stores features for everyone on the team to discover and share, reducing duplicative work.
Serves the same features for both training and inference, saving time and keeping features accurate
Analyzes and monitors features for drift.
Maintains a register of features with all their metadata and statistics, so that the whole team can work from a single source of truth.
Manages data for security and compliance purposes.
What are Features?
A feature is an input variable to a machine learning model. In other words, it’s a piece of data that will be consumed by a machine learning model. There are two types of ML features: online and offline.
Offline featuresare static features that don’t change often. This can be data like user language, location, or education level. These features are processed in batch. Typically, offline features are calculated via frameworks such as Spark, or by simply running SQL queries on a database and then using a batch inference process.
Online features—also called real-time features—are dynamic and require a processing engine to calculate, sometimes in real time. Number of ad impressions is a good example of a feature that changes very rapidly and would need to be calculated in real time. Online features often need to be served in ultra-low latency as well. For this reason, these calculations are much more challenging and require both speedy computation as well as fast data access. Data is stored in memory or in a very fast key-value database. The process itself can be performed on various services in the cloud or on a dedicated MLOps platform.
Why You Might Need a Feature Store
The data scientist’s strength is addressing business problems by understanding data and creating complex algorithms. They are not data engineers and they don’t need to be. In a typical workflow, data scientists search for and create features as part of their job, and the features they create are usually for training models in a strictly development environment. Thus, once the model is ready to be deployed in production, data engineers must take over and rewrite the feature to make it production-ready. This is a part of theMLOpsprocess (machine learning operationalization). This siloed process creates longer development cycles and introduces the risk of training-serving skew that could cause a less accurate model in production as a result of those code changes.
Real-time pipelines also require an extremely fast event processing mechanism while running complex algorithms to calculate features in real time. For many use cases in industries like Finance or AdTech, the application requires a response time in the range of milliseconds.
Meeting that requirement demands a suitable data architecture and the right set of tools to support real-time event processing with low-latency response times. ML teams cannot use the same tools for real-time processing as they do for training (e.g. Spark).
The key benefit of the feature store architecture is a very robust and fast data transformation service to power machine learning workloads, to address the challenges presented by data management and especially real-time data. A feature store solves the complex problem of real-time feature engineering, and maintains one logic for generating features for both training and serving. This way, ML teams can build it once and then use it for both offline training and online serving, ensuring that the features are being calculated in the same way for both layers, which is especially critical in low latency real time use cases.
Integrated or Stand-alone?
The feature store market is very active, with many new entrants over the past year and undoubtedly more to come. One of the most important characteristics of a feature store is that it is seamlessly integrated with other components in the ML workflow. Using anintegrated feature storewill make life simpler for everyone on the ML team, with monitoring, pipeline automation, and multiple deployment options already available, without the need for lots of glue logic and maintenance.
When it comes to the new world of analytics, the augmented analytics approach allows business users with no data science background to readily access and use analytics in an intuitive way. There are some important aspects of this approach, including auto machine learning, natural language processing (NLP) and intuitive search analytics.
Machine Learning via AutoML allows users to leverage systems and solutions that are designed with Machine Learning capabilities to predict outcomes and analyze data. Auto ML is the automated process of features and algorithm selection that supports planning, and allows users to fine tune, perform iterative modeling, and allows for the application and evolution of machine learning models. Machine Learning Algorithmsallows the system to understand data and applies correlation, classification, regression, or forecasting, or whichever technique is relevant, based upon the data the user wishes to analyze. Results are displayed using visualization types that provide the best fit for the data, and the interpretation is presented in simple natural language. This seamless, intuitive process enables business users to quickly and easily select and analyze data without guesswork or advanced skills.
With natural language-processing-based search capability, users do not need to scroll through menus and navigation. The business can address complex questions using this simple search capability with a contextual flexible search mechanism that provides one of the most flexible, in-depth search capability and results offered in the market today.
Clickless Analysis and contextual search capabilities go beyond column level filters and queries to provide more intelligence support and translate the contextual query and returns results in an appropriate format, e.g., visualization, tables, numbers, or descriptors. It takes natural language processing search analytics (NLP) and predictive modeling for business users to the next level and frees business users to produce accurate, clear results, quickly and dependably, using machine learning that frees the business user to collect and analyze data with the guided assistance of a ‘smart’ solution.
This foundation and these techniques come together to enable the enterprise and its business users to perform complex data analytics and share analysis across the organization in a self-serve, mobile environment. It brings the power of sophisticated, advanced analytics and smart data visualization to the next level with tools for automated data insights.
If you want to make encourage your business users to adopt and leverage the Clickless Analytics approach to NLP search analytics, and capitalize on intuitive Search Analytics and Auto Insights features that improve results and user adoption, Contact Us to get started. Read our Blog to find out more about Clickless Analytics and Natural Language Processing.
Can data be considered as the new gold? Considering the pace at which data is evolving all across the globe, there is little question. Consider the following:
Netflix saves $1 billion per year on customer retention only by utilizing big data.
Being the highest shareholder of the search engines market, Google faces1.2 trillionsearches every year, with more than 40,000 search queries every second!
Additionally, among all the google searches. 15% of those are new and are never typed before, leading to the fact that a new set of data is generated by Google continuously regularly. The main agenda is to convert data into information and then convert that information into insights.
Organizations were storing tons of their data into their databases without knowing what to do with that data until big data analytics became a completely developed idea. Poor data quality can cost businesses from $9.7 billion to14.2 millionevery year. Moreover, poor data quality can surely lead to wrong business strategies or poor decision-making. This also results in low productivity and sabotages the relationship between customers and the organization, causing the organization to lose its reputation in the market.
To deter this problem, here is a list of five things an enterprise must acquire in order to turn their big data into a big success:
Strong Leadership Driving Big Data Initiatives
The most important factor for nurturing data-driven decision-making culture is proper leadership. Organizations must have well-defined leadership roles for big data analytics to boost the successful implementation of big data initiatives. Necessary stewardship is crucial for organizations for making big data analytics an integral part of regular business operations.
Leadership-driven big data initiatives assist organizations in making their big data commercially viable. Unfortunately, only 34% of the organizations have appointed a chief data officer to handle the implementation of big data initiatives. A pioneer in the utilization of big data in the United States’s banking industry, Bank of America, specified a Chief Data Officer (CDO) who is responsible for all the data management standards and policies, simplification of It tools and infrastructures that are required for the implementation, and setting up thebig data platformof the bank.
Invest in Appropriate Skills Before Technology
Having the right skills are crucial even before the technology has been implemented:
utilize disparate open-source software for the integration and analysis of both structured and unstructured data.
framing and asking appropriate business questions with a crystal-clean line of sight such as how the insights will be utilized, and
bringing the appropriate statistical tools to bear on data for performing predictive analytics and generating forward-looking insights.
All of the above-mentioned skills can be proactively developed for both hiring and training. It is essential to search for those senior leaders within the organization who not only believe in the power of big data but are also willing to take risks and perform experimentation. Such leaders play a vital role in driving swift acquisitions and the success of data applications.
Perform Experimentation With Big Data Pilots
Start with the identification of the most critical problems of the business and how big data serves as the solution to that problem. After the identification of the problem, bring numerous aspects of big data into the laboratory where these pilots can be run before making any major investment in the technology. Such pilot programs provide an enormous collection of big data tools and expertise that prove value effectively for the organization without making any hefty investments in IT costs or talent. By working with such pilots, implementation of these efforts at a grassroots level can be done with minimal investments in the technology.
Search For a Needle in an Unstructured Hay
The thing that always remains on the top of the mind of businesses is unstructured and semistructured data – information contained in documents, spreadsheets, and similar non-traditional data sources. According to Gartner, data of organizations will evolve by 800% in the upcoming five years and 80% of that data will be unstructured. There are three crucial principles associated with unstructured data.
Having the appropriate technology is essential for storing and analyzing unstructured data.
Prioritiing such unstructured data that is rich in information value and sentiments.
Extracting relevant signals must be done from the insights and must be combined with structured data for boosting business predictions and insights.
Incorporate Operational Analytics Engines
One potential advantage that can be attained by using big data is the capability of tailoring experiences to customers based on their most up-to-the-minute behavior. Businesses can no longer extract the data of last month, analyze that data offline for two months, and act upon the analysis three months later for making big data a competitive benefit.
Take, as an example, loyal customers who enter promotional codes at the time of checkout but discover that their discount is not applied resulting in a poor customer experience.
Businesses need to shift their mindset of traditional offline analytics to tech-powered analytic engines that empower businesses with real-time and near-time decision-making, acquiring a measured test and learn approach. This can be achieved by making 20% of the organization’s decisions with tech-powered analytical engines and then gradually increasing the percentage of decisions processed in this way over time as comfort grows about the process..
Final Thoughts
In this tech-oriented world and digitally powered economy, big data analytics plays a vital role in the proper navigation of the market and to come up with appropriate predictions as well as decisions. Organizations must never ignore understanding patterns and deterring flows. especially as enterprises deal with different types of data each day, in different sizes, shapes, and forms. The market of big data analytics is growing dramatically, and will reach up to$62.10 billionby the year 2025. Considering that progression, 97.2% of the organizations are already investing in artificial intelligence as well as big data. Hence organizations must acquire appropriate measures and keep in mind all the crucial above-mentioned tips for turning their big data into big success to stay competitive in this ever-changing world.
Can data be considered as the new gold? Considering the pace at which data is evolving all across the globe, definitely yes!
Let me show you some eye-opening facts and statistics.
Do you know that Netflix saves $1 billion per year on customer retention only by utilizing big data? That’s not all. Being the highest shareholder of the search engines market, Google faces1.2 trillionsearches every year, with more than 40,000 search queries every second! There’s more, among all the google searches. 15% of those are new and are never typed before, leading to a new set of data generated by Google regularly. The main agenda is to convert data into information and then convert that information into insights.
Organizations were storing tons of their data into their databases without knowing what to do with that data until big data analytics became a completely developed idea. Poor data quality can cost businesses from $9.7 million to14.2 millionevery year. Moreover, poor data quality can surely lead to wrong business strategies or poor decision-making. This also results in low productivity and sabotages the relationship between customers and the organization, causing the organization to lose its reputation in the market.
To deter this problem, here is the list of 5 promising tips enterprises must acquire to turn their big data into a big success.
1. A Strong Leader for Driving Big Data Initiatives
The most important factor for nurturing data-driven decision-making culture is proper leadership. Organizations must have well-defined leadership roles for big data analytics to boost the successful implementation of big data initiatives. Necessary stewardship is crucial for organizations for making big data analytics an integral part of regular business operations. Leadership-driven big data initiatives assist organizations in converting their big data into a big success. Unfortunately, only 34% of the organizations have appointed a Chief Data Officer for the victorious implementation of big data initiatives. A pioneer in the utilization of big data in the United States’ banking industry, Bank of America, have specified a Chief Data Officer who is responsible for all the data management standards and policies, simplification of It tools and infrastructures that are required for the implementation, and setting up the big data platformof the bank.
2. Invest in Appropriate Skills Before Technology
Having the right technological skills is crucial, of which the following three are required:
The capability of utilizing disparate open-source software for the integration and analysis of both structured and unstructured data.
The capability of properly framing and asking appropriate business questions with a crystal-clean line of sight such as how the insights will be utilized.
The capability of bringing the appropriate statistical tools to bear on data for performing predictive analytics and generating forward-looking insights.
All of the above-mentioned skills can be proactively developed for both hiring and training. It is essential to search for those senior leaders within the organization who not only believe in the power of big data but are also willing to take risks and perform experimentation. Such leaders play a vital role in driving swift acquisitions and the success of data applications.
3. Perform Experimentation With Big Data Pilots
In the present age, numerous big data conversions emerge from technology vendors in case they have anything to do with the business case and return of investment (ROI) of big data. Start with the identification of the most critical problems of the business and how big data serves as the solution to that problem. After the identification of the problem, bring numerous aspects of big data into the big data laboratory where these pilots can be run before making any major investment in the technology. Big data labs provide an enormous collection of big data tools and expertise that permits organizations to run a pilot and prove value effectively without making any hefty investments in IT and talent. Implementation of these efforts at a grassroots level can be done with minimal investments in the technology.
4. Search For a Needle in an Unstructured Hay
The thing that always remains on the top of the mind of businesses is unstructured and semi-structured data. According to Gartner, data of organizations will grow by 800% in the upcoming five years and 80% of that data will be unstructured. Let us see the three most crucial principles associated with unstructured data.
Assurance of having the appropriate technology is essential for storing and analyzing unstructured data.
Prioritization and attention to such unstructured data are important that can be linked back to the individual. Also, it is imperative to prioritize such unstructured data that is rich in information value and sentiment.
Only analyzing the unstructured data is not enough. Extraction of relevant signals must be done from the insights and must be combined with structured data for boosting business predictions and insights.
5. Incorporate Operational Analytics Engines
One of the potential advantages that can be attained by using big data is the capability of tailoring experiences to customers based on their most up-to-the-minute behavior. Businesses can no longer extract the data of last month, analyze that data offline for two months, and act upon the analysis three months later for making big data a competitive benefit. Take the high-value case under consideration, loyal customers who enter the promotional code at the time of checkout but discount is not applied, resulting in poor customer experience. It’s high time for businesses to shift their mindset of traditional offline analytics to tech-powered analytic engines that empower businesses with real-time and near-time decision-making. Companies must acquire a measured test and learn approach. Making 20% of the organization’s decisions with tech-powered analytical engines and then gradually increasing the percentage of decisions help organizations in developing a greater level of comfort.
Final Thoughts
In this tech-oriented world and digitally powered economy, big data analytics plays a vital role in the proper navigation of the market and to come up with appropriate predictions as well as decisions. Organizations must never ignore at any cost the natural instincts of understanding patterns and deterring flows. Enterprises deal with different types of data each day. That data exists in different sizes, shapes, and forms. The market of big data analytics is tremendously progressing and will reach up to$62.10 billionby the year 2025. Considering that progression, 97.2% of the organizations are already investing in artificial intelligence as well as big data. Hence organizations must acquire appropriate measures and keep in mind all the crucial above-mentioned tips for turning their big data into big success to stay competitive in this ever-changing world.
By 2035 AI could boost average profitability rates by 38 percent and lead to an economic increase of $14 Trillion.
The words Artificial Intelligence (AI), and algorithms are most often misused and misunderstood. There are often used interchangeably when they shouldn’t be. This leads to unnecessary confusion.
In this article, let’s understand what AI and algorithms are, and what the difference between them is.
An algorithm is a form of automated instruction. An algorithm can either be a sequence of simple single if-then statements like if this button is pressed, execute that action, or sometimes it can be more complex mathematical equations.
· Examples where algorithms are used
YouTube’s algorithm knows what kind of ads should be displayed to a particular user
The e-commerce giant Amazon’s algorithm knows what kind of products a specific user like and based on it shows similar product details.
·Types of algorithms
The complexity of an algorithm will depend on the complexity of every single step, which is required to execute, as well as on the sheer number of steps the algorithm is required to execute. Mostly the algorithms are quite simpler.
Basic algorithm
If a defined input leads to a defined output, then the system’s journey can be called an algorithm. This program journey between the start and the end emulates the basic calculative ability behind formulaic decision-making.
Complex algorithm
If a system is able to come to a defined output based on a set of complex rules, calculations, or problem-solving operations, then that system’s journey can be called a complex algorithm. Same as the basic algorithm, this program journey emulates the calculative ability behind formulaic, but more complex decision-making.
Artificial intelligence is a set of algorithms, which is able to cope with unforeseen circumstances. It differs from Machine Learning (ML) in that it can be fed unstructured data and still function. One of the reasons why AI is often used interchangeably with ML is because it’s not always straightforward to know whether the underlying data is structured or unstructured. This is not so much about supervised and unsupervised learning, but about the way, it’s formatted and presented to the AI algorithm.
The term AI algorithms are usually used to mention the details of the algorithms. But the accurate word to use for this is “Machine Learning Algorithms”. AI is a culmination of technologies that embrace Machine Learning (ML). ML is a set of algorithms that enables computers to learn from previous outcomes and get an update with the information without human intervention. It is simply fed with a huge amount of structured data in order to complete a task.
Based on the data acquired, AI algorithms will develop assumptions and come up with possible new outcomes by considering several factors into account that help them to make better decisions than humans.
In AI algorithms, outputs are not defined but designated depending on the complex mapping of user data that is then multiplied with each output. This program’s journey emulates the human ability to come to a decision, based on collected data. The more an intelligent system can enhance its output based on additional inputs, the more advanced the application of AI becomes.
· Examples where AI algorithms are used
Self-driving cars are one of the best examples of AI algorithms.
Recognition-based applications such as facial, speech, and object recognition mapping
·Learning algorithms
Artificial intelligence algorithms are also called learning algorithms. There are three major kinds of algorithms in ML.
Supervised learning
The supervised learning algorithms are based on outcome and target variable mostly dependent variable. This gets predicted from a specific set of predictors which are independent variables. By making use of this set of variables, one can generate a function that maps inputs to get adequate results. The core algorithms, which are available in supervised learning, are Support Vector Machines (SVM), Decision Tree, and naïve Bayes classifiers, Ordinary Least Squares (OLS), Random Forest, Regression, Logistic Regression, and KNN.
Unsupervised learning
These are similar to the supervised learning algorithms, but there is no specific target or result, which can be estimated or predicted. As they keep on adjusting their models entirely based on input data. The algorithm operates a self-training process without any type of external intervention. The core algorithms, which are available in unsupervised learning algorithms, are Independent Component Analysis (ICA), apriori algorithm, K-means, Singular Value Decomposition (SVD), and Principal Component Analysis (PCA).
Reinforcement Learning (RL)
The RL has the constant iteration that depends on trial and error, in which the machines can generate the outputs depending on the specific kind of conditions, the machines are well-trained to take relevant decisions. The machine learns well based on past experiences and then captures the most suitable and relevant information to develop business decisions accurately. The best examples for RL are Q-Learning, Markov Decision Process, SARSA (State – action – reward – state – action), and Deep Mind’s Alpha Zero chess AI.
An algorithm takes automated instructions, which can be simple or complex, takes some input and some logic in the form of code, and offers an output based on the predefined set of guidelines described in the algorithm.
Whereas, an AI algorithm varies based on the data it receives whether structured or unstructured learns from the data and comes up with unique solutions. It also possesses the capability to alter its algorithms and develop new algorithms in response to learned inputs.
Humans and machines must work together to build humanized technology grounded by diverse socio-economic backgrounds, cultures, and various other perspectives. Knowledge of algorithms and AI will help to develop better solutions and to be successful in today’s volatile and complex world.
Before launching into my editorial this week, I wanted to make the announcement that starting with this issue, the DSC Newsletter will be sent out on Tuesdays rather than Mondays. To subscribe to the DSC Newsletter, go to Data Science Central and become a member today. It’s free!
There has long been a pattern with computer technology. At a certain point in the evolution of technology, there is a realization that things that had been done repeatedly as one-offs occur often enough to start building libraries, or even extensions to languages. For a while, mastery of these libraries defines a certain subset of programmers. or analysts, and typically most of the innovations tend to take place as improvements to these libraries, articles on technical sites or journals, and so forth.
Eventually, however, the capabilities are abstracted into more sophisticated stand-alone applications, frequently with user interfaces that provide ways to handle the most frequent use cases while relegating the edge cases to specialized screens. The results of these, in turn, are wrapped within some kind of container (such as Kubernetes) that can then be incorporated into a pipeline with other similar containers.
This is the direction that machine learning is going. MLOps now complements DevOps, ensuring that changes to machine learning models – from the data engineering necessary to ensure that the source data is ready for production, to feature engineering that can be altered on the fly to try out different scenarios, through to presentation and productization that not only makes sure that the results are understandable to a business audience, but that also can then feed into other operational channels, is here now, and will likely become commonplace within the next couple of years within the industry.
This transformation is critical for several reasons. First, it makes it far easier to create ensemble models, models that are developed and work in parallel, and that can handle different starting scenarios. This is key because the more generalized a model has to be, the more expensive, time-consuming, and complex it turns out, and the less likely that it can handle edge cases accurately. This is especially important when dealing with sparse datasets, where the danger is that single, comprehensive models can badly overfit the input, making such models very brittle to initial conditions.
In addition to this, however, however, by reducing the overall costs of implementing models from months to weeks, or even days, organizations are able to better productize their data analytics in ways that would have been unheard of even a couple of years before. As not all problems can (or should) be solved with machine learning in the first place, the ability to take advantage of more generalized DevOps pipelines within your organization put machine learning right where it belongs – as a powerful tool among many, rather than a single, potentially shaky foundation on its own.
For machine learning and data science specialists, this has other implications as well. Domain proficiency in a given sector will mean more, the ability to write Python or R will mean less, save for those who focus more specifically on tool-building within integrated frameworks. However, having a good understanding of data operations in general and machine learning operations in particular, all engineering tasks, will likely increase in demand dramatically over the next few years. Additionally, those that are better at productizing data, integrating ML streams in with other streams towards the creation of digital assets that can then be published as physical assets, will do quite well.
Machine learning is maturing. There’s nothing wrong with that.
