Analyzing E-commerce Customer Behavior for Churn Prediction

Overview

This project focuses on predicting customer churn within an e-commerce platform using advanced data mining techniques. Understanding customer behavior is pivotal in today's digital marketplace to enhance retention strategies and address churn. The primary goal is to explore an e-commerce dataset, predict churn, and derive actionable insights for better customer retention.

Dataset

Source: Kaggle link
Size: 5630 observations, 20 attributes
Nature: Mix of categorical, numerical, and textual attributes capturing various facets of customer behavior and transactional information.

Key Attributes

CustomerID: Unique identifier
Churn: Target variable (1 for churned, 0 for active)
Other attributes like Tenure, PreferredLoginDevice, PreferredPaymentMode, SatisfactionScore, etc., crucial for understanding customer behavior and predicting churn.

Methodology

Talend Data Preparation: Exploratory analysis of dataset attributes to understand their impact on churn prediction.
Talend Data Integration: Preprocessing, handling missing values, dropping irrelevant attributes, and transforming categorical variables for modeling.
SAS Enterprise Miner: Building decision trees, Random Forest, and Gradient Boost models for churn prediction and comparison.

Results & Insights

Model Comparison: Decision Tree emerged as the most accurate for predicting churn, surpassing Random Forest and Gradient Boost.
Key Predictors: Attributes like Tenure, Complain, PreferredPaymentMode, among others, significantly influence churn behavior.
Implications: Insights aid in enhancing customer retention strategies, targeted marketing, and improving service quality.

Usage

Clone the repository and follow instructions in each tool's respective folder for replicating data preparation, preprocessing, and modeling steps.

For more examples, please refer to the Documentation.

Folder Structure

.
├── WQD7005_AA1_Analysis # Contains SAS e-Miner analysis & modeling scripts.
├── WQD7005_AA1_Data_Integration # Includes Talend Data Integration preprocessing scripts.
├── WQD7005_Report
├───── Assessment.pdf # The assessment requirement file of this project.
├───── Report.pdf # The documentation (report) file of this project.
├── dataset
├───── E Commerce Dataset.xlsx # Raw dataset in Excel format.
├───── E Commerce Dataset.csv # Raw dataset in CSV format.
├───── E Commerce Dataset (processed).csv # Preprocessed dataset in CSV format.

Contact

Name: LOO JIA HAO
Student No: 17218692/1
Student Email: 17218692 (at) siswa (dot) um (dot) edu (dot) my

Feel free to explore each folder for detailed scripts, reports, and documentation related to specific phases of the project.

jvloo/WQD7005_AA1