Building a Customer Churn Prediction System The project focuses on creating a machine learning model to predict customer churn using the Kaggle Telecom dataset. Churn refers to customers discontinuing their subscription or service. The workflow includes data collection, exploratory data analysis (EDA), preprocessing, and training tree-based models like Random Forest Classifier.
Data Collection and Initial Setup A CSV file from Kaggle is used as the dataset containing features such as gender, tenure, monthly charges, etc., with 'churn' being the target column. Data is loaded into pandas for initial inspection of rows/columns count and missing values identification.
Exploratory Data Analysis Importance EDA helps understand datasets better by identifying necessary preprocessing steps like handling class imbalance or selecting suitable models through visualization techniques that reveal patterns in numerical/categorical columns.
Preprocessing Steps Overview 'Churn' column encoding involves replacing categorical strings ('yes', 'no') with binary integers (1s & 0s). Label encoders are applied across other non-numeric fields ensuring compatibility during modeling stages while addressing imbalances via SMOTE oversampling technique post train-test split operations