Delete models.md

models.md

@@ -1,37 +0,0 @@
-# Comparison of Training Models
-
-1. **Lite Model**
-- Uses basic ensemble voting with fixed weights [2,1,2]
-- Pre-configured hyperparameters (no optimization)
-- Lemmatization for text preprocessing
-- Uses 3 algorithms: SVC, MultinomialNB, ExtraTreesClassifier
-- Fastest because no parameter tuning/optimization
-
-2. **Legacy Model**
-- Uses simple voting ensemble without weight optimization
-- Porter Stemming for text preprocessing (simpler than lemmatization)
-- Slightly different hyperparameters:
-  - SVC with 'sigmoid' kernel
-  - Fewer trees in ExtraTreesClassifier (50 vs 200)
-- Medium speed due to simpler preprocessing
-
-3. **Monarch Butterfly Optimization (MBO) Model**
-- Uses nature-inspired optimization algorithm
-- Optimizes 7 parameters simultaneously:
-  - SVC parameters (C, gamma)
-  - MultinomialNB alpha
-  - Number of trees
-  - Ensemble weights (w1, w2, w3)
-- Population-based search with:
-  - 20 butterflies
-  - 30 iterations
-  - Cross-validation for each evaluation
-- Slowest because:
-  - Runs multiple training cycles (20 butterflies × 30 iterations)
-  - Each evaluation requires 5-fold cross-validation
-  - Total of ~3000 model evaluations
-
-**Summary**:
-- Lite: Quick, fixed parameters
-- Legacy: Traditional, basic ensemble
-- MBO: Advanced optimization, but computationally intensive