// case study
Comparing ML models on nutrition data
Compared 18 model and scaler combinations; best accuracy 0.8975 (MLP).
Context
An artificial intelligence project from my previous portfolio, where it was titled "Model Clustering".
What I did
- Correlation analysis of nutrition features (calories, proteins, fat, carbohydrate) with a heatmap. The strongest correlation was between calories and fat (0.69).
- Compared MLP, SVC, KNN, Random Forest, and Decision Tree (Gini/Entropy), each with no scaler, StandardScaler, and MinMaxScaler.
- Measured accuracy, precision, recall, and F1 score.
Result
MLP with MinMaxScaler scored the highest accuracy, 0.8975 (F1 0.8956). Decision Tree Entropy without a scaler was lowest at 0.8125.