This project involves predicting customer churn in a telecommunications company using machine learning techniques, exploring various features' impact, optimizing models, and identifying key factors influencing churn.
In this Case Study we will be working for a fintech company that wants to provide its customers with a paid mobile app subscription that will allow them to track all of their finances in one place. To attract customers, the company releases a free version of their app with some of the main features unlocked.
The company has tasked you to identify which users will most likely not enroll in the paid product, so that additional offers can be given to them. Because of the costs of these offers, the company does not want to offer them to everybody, especially customers who were going to enroll anyways.
-
Market: The target audience is customers who use a company's free product. In this case study, this refers to users who installed (and used) the company's free mobile app.
-
Product: The paid memberships often provide enhanced versions of the free products already given for free, alongside new features.
-
Goal: The objective of this model is to predict which users will not subscribe to the paid membership, so that greater marketing efforts can go into trying to "convert" them to paid users.
By working for the company, we have access to each customer's app behavior data. This data allows us to see the date and time of app installation, as well as the features the users engaged with within the app. App behavior is characterized as the list of app screens the user looked at, and whether the user played the financial mini-games available.
The app usage data is only from the user's first day in the app. This limitation exists because users can enjoy a 24-hour free trial of the premium features, and the company wants to target them with new offers shortly after the trial is over.
Above mentioned Dataset is stored in file - "appdata10.csv"
Luckily Company also provided list of top screen used by user to reduce number of feature in file - "top_screens.csv"
"fintech.py" and "fintech.ipynb" are python and notebook files respectively.
- Statistical Analysis of numerical variable with dependent variable and correlation matrix.
- Data Cleaning and Analysis based on difference between first_open and enrolled_time.
- Feature Extraction from "screen_list" field to several new field listed in "top_screens".
- Training Logistic Regression with hyperparameter tuning.
- Training XGBoost Model and performance comparision based on Classification Metrics such as confusion_matrix, accuracy, precision, recall, f1_score.