Customer Classification in E-Commerce Using Random Forest Algorithm

الملخص

This study investigates how well Random Forest Classification machine learning algorithms predict customer satisfaction levels when shopping online. A synthetic dataset of 350 customer records containing demographics, behavior and transaction data was assembled with the purpose of segmenting customers into three categories of satisfaction (Neutral/Satisfied/Unsatisfied). All data was divided into training and test datasets using an 80/20 stratified split, and all data was preprocessed and transformed into numerical codes. Different evaluation metrics were used for evaluating performance; these include Accuracy, Precision, Recall, F1 Score, Confusion Matrix, Cross Validation, and Multiclass ROC-AUC. The Random Forest algorithm outperformed the other algorithms with a total accuracy rate of 98.57%, a mean Cross Validation accuracy rate of 98.86%, and a mean ROC-AUC score of 0.9963. The feature importance analysis indicated that Days Since Last Purchase and Total Spend are the most important factors influencing customer satisfaction. The Random Forest algorithm has shown superior performance with large and complex e-commerce datasets, which can help retailers improve how they retain customers and create customized marketing programs. This study is part of an increasing body of knowledge aimed at using data to improve the quality of customer satisfaction forecasting for online retailers.