Data analysis and machine learning to predict and understand whether a customer will churn or not.
In this project, the dataset I have analysed contains information about customers of a bank. I started off with EDA (Exploratory Data Analysis) to visualize and extract key information that influence customers to churn. Then I delved into building Machine Learning models to try and predict whether a customer would churn or not. This is a classification problem, and for evaluating the models and optising them, the performance metric I used is Recall. Comparisons using the other metrics have been showed as well. I have chosen Recall since its more important for banks to correctly identify the positive class (true churn).
At the end of this project, I was able to come up with some key findings and suggestions, which the bank could use to improve their customer retention. In addition, the findings could help them understand their customers better.
Skills demonstrated: Exploratory Data Analysis, Visualisation, Data Preprocessing (Feature Selection, Feature Encoding, Feature Scaling), Dealing with Class Imbalance using SMOTE, and Model Tuning.
ML models used: Logistic Regression, Support Vector Machines, Random Forests, Gradient Boosting, and XGBoost.
