Published Paper Details:

Customer churn prediction is crucial for e-commerce companies looking to improve retention and optimize business strategies. In this study, we aim to build a robust predictive model to identify customers at risk of churn, using a variety of customer features such as demographics, behavior metrics, order history, and satisfaction scores. The analysis begins with an exploratory data analysis (EDA), using Python libraries like pandas, numpy, matplotlib, and seaborn for data manipulation, visualization, and understanding patterns in the dataset. Missing values are visualized and addressed using missing no, ensuring data quality before model development. Several machine learning models are trained to predict customer churn, including Support Vector Classifier (SVC), Logistic Regression, Random Forest Classifier, and XGBoost, implemented with libraries such as scikit-learn and xgboost. Data preprocessing steps, such as scaling features with Standard Scaler and encoding categorical variables using LabelEncoder, are performed to prepare the dataset for model training. Model evaluation is conducted using performance metrics like accuracy, precision, recall, F1-score, ROC-AUC, and confusion matrix, offering detailed insights into the effectiveness of each model. In addition, hyperparameter tuning is achieved using GridSearchCV to optimize model performance. Cross-validation is applied with cross validate to ensure generalizability and robustness of the models. The study identifies key factors influencing churn, including marital status, order categories, and platform engagement, and provides actionable insights for the company to implement targeted retention strategies. By leveraging machine learning and data analytics, this research helps the e-commerce company proactively manage churn, enhance customer satisfaction, and optimize product offerings tailored to customer preferences.