Predictive Modeling for Food Delivery Customer Ratings
Jul 28, 2025
What I Did:
- Built and compared 4 predictive models (CART, ANN, CHAID, RF) to identify key drivers of high customer ratings (≥4 stars) for food delivery services.
- Cleaned and preprocessed data (3,639 → 2,779 records), handled class imbalance with SMOTE, and optimized model performance.
Key Results:
- CHAID model achieved 86.3% accuracy, revealing ASAP pickup availability and delivery speed as top predictors.
- Actionable insights: Restaurants with ASAP pickup had 82.3% confidence of high ratings.
Tools: IBM SPSS Modeler, Excel | Skills: Predictive Modeling, Data Cleaning, Feature Importance