Brazilian e-Commerce Transaction 5-star Review Classifier

This project purpose it to develop ML model to classify whether a transaction will resulted in 5-star review or not. This is part of final evaluation for Sharing Vision Data Science Bootcamp that I've done.

Model Success Criteria

• Recall > 0.7
• Precision > 0.7
• FPR < 0.3

What to be Seen

• Data preprocessing using Winsorization, Square & Log transformation, OHE, frequency encoding, mutual info feature selection
• Classification model inspected : DT, RF, AdaBoost, XGBoost, SVM.
• Model tuning methods : GridSearch, Sequential, Threshold adjustment

Datasets

The datasets used were given by bootcamp coordinators, however I also found several datasets which may or may not be used by this dataset as source. Kaggle