🏆 AI-ML Competition - E-Commerce Customer Purchase Prediction

📌 Repository Overview

Welcome to my AI-ML Competition Repository! 🚀 This project focuses on predicting customer purchases using machine learning techniques. The goal is to analyze customer data, apply feature engineering, and train an AI model to forecast whether a customer will make their next purchase.

This repository includes data preprocessing, model training, evaluation, and key insights for improving customer targeting strategies. 📊

---

🔬 1. Methodology

📌 Data Preprocessing

• 🛠️ Handling missing values in Annual_Income by imputing median values.
• 📆 Transforming Birth_Year into Age for better feature representation.
• 🏷️ Encoding categorical variables (Education Level, Family Status) using One-Hot Encoding.
• 📏 Normalizing numerical features for consistent scaling.
• 📂 Splitting dataset: 80% Training / 20% Validation.

---

🤖 2. Model & Performance Metrics

A Random Forest Classifier was trained to predict customer purchases.

📊 Performance Metrics

✅ Accuracy: 86.14% 🎯
✅ F1-Score: 0.53 (indicating room for improvement in predicting buyers)
✅ Precision & Recall: The model performs well for non-buyers, but struggles with buyers.

🔍 Confusion Matrix

• The model correctly identifies most non-buyers, but has a higher false negative rate for buyers.

📈 ROC Curve & AUC Score

• ROC Curve demonstrates the model’s class distinction ability.
• AUC Score suggests moderate performance, requiring further improvements.

💡 Feature Importance

• Key factors influencing purchases:
  • Spending habits 💰
  • Promotional campaigns 🎯
  • Past purchases 🛍️

---

📊 3. Business Insights & Recommendations

🔍 Key Insights

• 📌 Customers with prior purchases are more likely to buy again.
• 🏷️ Promotional campaigns play a major role in repeat purchases.
• 💰 High-income customers tend to purchase more frequently.
• ❗ The model struggles with rare buyers (Class 1) → Need for data balancing.

📌 Recommendations

⚖️ Balance dataset using SMOTE to improve prediction for rare buyers.
🛠️ Hyperparameter tuning (adjust n_estimators, max_depth) to enhance performance.
🚀 Explore alternative models like XGBoost or Neural Networks for better accuracy.
🎯 Optimize marketing campaigns by targeting customers more likely to purchase.

---

🏆 Final Thoughts

This predictive model provides valuable insights to enhance marketing strategies and boost customer engagement. By implementing the suggested improvements, the model can achieve higher accuracy, better F1-scores, and more effective business decisions. 🚀

🔗 Follow this repository for updates and improvements in AI-ML predictions! 🎯