Day 2 - Domain 3 Integration: Modeling (Algorithm to Evaluation)

[Business problem]
   │
   ├─ 1) DEFINE ML TYPE
   │     Label? → supervised (classification/regression) or unsupervised
   │     Output? → category/number/cluster/anomaly/recommendation
   │
   ├─ 2) PICK ALGORITHM
   │     Tabular → XGBoost (default)
   │     Image → CNN
   │     Sequence/TS → RNN/LSTM or DeepAR
   │     Text → BlazingText
   │
   ├─ 3) STRUCTURE (if DL)
   │     Activate: ReLU (hidden) → Sigmoid/Softmax/linear (output)
   │     Loss: cross-entropy (classify) or MSE (regress)
   │
   ├─ 4) OPTIMIZE/TUNE
   │     Hyperparameter sear