Merge gh pages into main

GNN/RESULTS_SUMMARY.md

@@ -24,7 +24,7 @@ The method produces spatially coherent hotspot regions that capture all high-sev
 
 ---
 
-## Approach: Hybrid Spatial Modeling + Clustering
+### Performance Winner: Hybrid GNN Modeling + Clustering
 
 This approach integrates multiple complementary components.
 

GNN/src/baseline_comparison.py

@@ -157,9 +157,7 @@ def compare_all_baselines(self, metadata: pd.DataFrame,
         Returns:
             Comparison results dictionary
         """
-        print("\n" + "="*60)
-        print("Running Baseline Comparisons")
-        print("="*60)
+        print("\nBaseline Comparison")
 
         results = {}
 

GNN/src/data_split.py

@@ -197,9 +197,7 @@ def print_split_summary(self, splits: Dict):
         Args:
             splits: Dictionary with split data
         """
-        print("\n" + "="*50)
-        print("Data Split Summary")
-        print("="*50)
+        print("\nData Split Summary")
 
         if 'train' in splits:
             if isinstance(splits['train'], dict):
@@ -218,5 +216,5 @@ def print_split_summary(self, splits: Dict):
                         pct = (n / total) * 100 if total > 0 else 0
                         print(f"{split_name.capitalize()}: {n} samples ({pct:.1f}%)")
 
-        print("="*50 + "\n")
+        print("")
 

GNN/src/evaluate.py

@@ -179,9 +179,7 @@ def print_metrics(self, metrics: Dict):
         Args:
             metrics: Dictionary of metrics
         """
-        print("\n" + "="*50)
-        print("Evaluation Metrics")
-        print("="*50)
+        print("\nEvaluation Metrics")
 
         if 'silhouette_score' in metrics:
             print(f"Silhouette Score: {metrics['silhouette_score']:.4f}")
@@ -200,5 +198,5 @@ def print_metrics(self, metrics: Dict):
         print(f"  Hotspot ratio: {metrics.get('hotspot_ratio', 0):.2%}")
         print(f"  Avg risk in hotspots: {metrics.get('avg_risk_in_hotspots', 0):.4f}")
         print(f"  Avg risk outside: {metrics.get('avg_risk_outside_hotspots', 0):.4f}")
-        print("="*50 + "\n")
+        print("")
 

GNN/src/main.py

@@ -59,9 +59,7 @@ def main():
     print(f"Using device: {device}")
 
     # 1. Data Preprocessing
-    print("\n" + "="*50)
-    print("Step 1: Data Preprocessing")
-    print("="*50)
+    print("\nStep 1: Data Preprocessing")
     preprocessor = DataPreprocessor(config)
     features, metadata = preprocessor.process()
     print(f"Preprocessed {len(features)} data points")
@@ -70,24 +68,18 @@ def main():
     coords = metadata[['lon', 'lat']].values
 
     # 2. Graph Construction
-    print("\n" + "="*50)
-    print("Step 2: Graph Construction")
-    print("="*50)
+    print("\nStep 2: Graph Construction")
     graph_builder = GraphBuilder(config)
     graph_data = graph_builder.build_graph(coords, features, metadata)
     print(f"Graph constructed: {graph_data.num_nodes} nodes, {graph_data.num_edges} edges")
 
     # 2.5. Data Splitting
-    print("\n" + "="*50)
-    print("Step 2.5: Data Splitting")
-    print("="*50)
+    print("\nStep 2.5: Data Splitting")
     data_splitter = DataSplitter(config)
     graph_splits, metadata_splits = data_splitter.split_graph_data(graph_data, metadata)
 
     # 3. Model Setup
-    print("\n" + "="*50)
-    print("Step 3: Model Setup")
-    print("="*50)
+    print("\nStep 3: Model Setup")
     input_dim = features.shape[1]
     model = GNNModel(input_dim, config['model'])
     print(f"Model created: {config['model']['architecture']}")
@@ -99,9 +91,8 @@ def main():
     model_path = Path(__file__).parent.parent / model_path
 
     if args.train or not model_path.exists():
-        print("\n" + "="*50)
-        print("Step 4: Training")
-        print("="*50)
+        # 4. Training
+        print("\nStep 4: Training")
         history = trainer.train(
             graph_splits['train'], 
             metadata_splits['train'],
@@ -115,9 +106,7 @@ def main():
         trainer.load_model(str(model_path))
 
     # 5. Feature Extraction (on test set for evaluation)
-    print("\n" + "="*50)
-    print("Step 5: Feature Extraction & Aggregation")
-    print("="*50)
+    print("\nStep 5: Feature Extraction & Aggregation")
     feature_extractor = FeatureExtractor(model, config, device)
 
     # Extract embeddings on test set for evaluation
@@ -136,19 +125,15 @@ def main():
     print(f"Extracted and aggregated embeddings: {aggregated_embeddings.shape} (test set)")
 
     # 6. Hotspot Scoring (on test set)
-    print("\n" + "="*50)
-    print("Step 6: Hotspot Scoring (Test Set)")
-    print("="*50)
+    print("\nStep 6: Hotspot Scoring (Test Set)")
     hotspot_scorer = HotspotScorer(config)
     risk_scores, score_components = hotspot_scorer.score(
         metadata_splits['test'], aggregation_metadata, aggregated_embeddings
     )
     print(f"Computed risk scores for {len(risk_scores)} spatial units")
 
     # 7. Hotspot Detection (on test set)
-    print("\n" + "="*50)
-    print("Step 7: Hotspot Detection (Test Set)")
-    print("="*50)
+    print("\nStep 7: Hotspot Detection (Test Set)")
     hotspot_detector = HotspotDetector(config)
     aggregation_coords = aggregation_metadata[['center_lon', 'center_lat']].values
     hotspot_results = hotspot_detector.detect_hotspots(
@@ -160,9 +145,8 @@ def main():
     # 8. Evaluation (on test set)
     metrics = {}
     if args.eval:
-        print("\n" + "="*50)
-        print("Step 8: Evaluation (Test Set)")
-        print("="*50)
+        # 8. Evaluation (on test set)
+        print("\nStep 8: Evaluation (Test Set)")
         evaluator = Evaluator(config)
         metrics = evaluator.evaluate(
             aggregated_embeddings,
@@ -214,9 +198,7 @@ def convert_to_native(obj):
 
     # 9. Visualization (on full data for better visualization)
     if args.visualize:
-        print("\n" + "="*50)
-        print("Step 9: Visualization (Full Data)")
-        print("="*50)
+        print("\nStep 9: Visualization (Full Data)")
         output_dir = Path(config['paths']['output_dir'])
         output_dir = Path(__file__).parent.parent / output_dir
         visualizer = Visualizer(config, str(output_dir))
@@ -236,9 +218,8 @@ def convert_to_native(obj):
 
     # 10. Create Dashboards
     if args.eval:
-        print("\n" + "="*50)
-        print("Step 10: Creating Dashboards")
-        print("="*50)
+        # 10. Create Dashboards
+        print("\nStep 10: Creating Dashboards")
         dashboard_dir = Path(__file__).parent.parent / config['paths']['output_dir'] / 'dashboards'
         dashboard = ModelDashboard(str(dashboard_dir))
 
@@ -254,9 +235,8 @@ def convert_to_native(obj):
 
     # 11. Baseline Comparison
     if args.eval:
-        print("\n" + "="*50)
-        print("Step 11: Baseline Comparison")
-        print("="*50)
+        # 11. Baseline Comparison
+        print("\nStep 11: Baseline Comparison")
         baseline_comparison = BaselineComparison(config)
 
         # Prepare GNN results (using test set results)
@@ -318,9 +298,7 @@ def convert_for_json(obj):
         from .update_results_with_baselines import update_results_summary_with_baselines
         update_results_summary_with_baselines()
 
-    print("\n" + "="*50)
-    print("Pipeline Complete!")
-    print("="*50)
+    print("\nPipeline Complete!")
 
 
 if __name__ == '__main__':

temporal_calc/config.py

@@ -5,9 +5,7 @@
 import os
 from pathlib import Path
 
-# ============================================================================
 # Paths
-# ============================================================================
 BASE_DIR = Path(__file__).parent
 DATA_DIR = BASE_DIR.parent / "data"
 OUTPUT_DIR = BASE_DIR / "outputs"
@@ -23,9 +21,7 @@
 for dir_path in [OUTPUT_DIR, PLOTS_DIR, EVAL_DIR, CHECKPOINT_DIR]:
     dir_path.mkdir(parents=True, exist_ok=True)
 
-# ============================================================================
 # Data Configuration
-# ============================================================================
 # Number of synthetic time bins to create from attribute_id ordering
 # Higher = more granular time series, more training data per neighborhood
 NUM_TIME_BINS = 200  # Increased from 50 for denser data
@@ -55,9 +51,7 @@
     "Other": 1.0
 }
 
-# ============================================================================
 # Model Configuration (Optimized for M1 Mac 8GB RAM)
-# ============================================================================
 MODEL_CONFIG = {
     "d_model": 128,          # Embedding dimension (increased from 64)
     "n_heads": 8,            # Number of attention heads (increased from 4)
@@ -67,9 +61,7 @@
     "max_seq_len": 30,       # Maximum sequence length
 }
 
-# ============================================================================
 # Training Configuration
-# ============================================================================
 TRAIN_CONFIG = {
     "seq_len": 8,            # Input sequence length (reduced for more samples)
     "pred_len": 1,           # Prediction horizon
@@ -81,9 +73,7 @@
     "seed": 42,
 }
 
-# ============================================================================
 # Device Configuration
-# ============================================================================
 def get_device():
     """Get the best available device (MPS for M1 Mac, else CPU)."""
     import torch

temporal_calc/max_performance.py

@@ -78,7 +78,7 @@ def calculate_transformer_metrics(preds, actuals, tolerance=1.0):
 print(f"Using device: {DEVICE}")
 
 
-# ============= DATA AUGMENTATION =============
+# DATA AUGMENTATION
 
 class AugmentedDataset(Dataset):
     """Dataset with noise injection and temporal jittering."""
@@ -140,7 +140,7 @@ def eval(self):
         self._training = False
 
 
-# ============= ENHANCED TRANSFORMER =============
+# ENHANCED TRANSFORMER
 
 class MaxPerformanceTransformer(nn.Module):
     """
@@ -223,7 +223,7 @@ def forward(self, x):
         return self.output_head(cls_output)
 
 
-# ============= BIDIRECTIONAL LSTM =============
+# BIDIRECTIONAL LSTM
 
 class BiLSTMModel(nn.Module):
     """
@@ -273,7 +273,7 @@ def forward(self, x):
         return self.output_head(context)
 
 
-# ============= LEARNING RATE SCHEDULER =============
+# LEARNING RATE SCHEDULER
 
 class CosineWarmupScheduler:
     """Cosine annealing with linear warmup."""
@@ -304,7 +304,7 @@ def get_lr(self):
         return self.optimizer.param_groups[0]['lr']
 
 
-# ============= TRAINING =============
+# TRAINING
 
 def train_model(model, train_loader, val_loader, target_scaler,
                 model_name: str, epochs: int = 200, lr: float = 3e-4,

temporal_calc/optimized_training.py

@@ -46,7 +46,7 @@
 print(f"Using device: {DEVICE}")
 
 
-# ============= 1. ENHANCED FEATURE ENGINEERING =============
+# 1. ENHANCED FEATURE ENGINEERING
 
 def add_enhanced_temporal_features(df: pd.DataFrame, target_col: str = 'accessibility_score') -> pd.DataFrame:
     """
@@ -114,7 +114,7 @@ def add_enhanced_temporal_features(df: pd.DataFrame, target_col: str = 'accessib
     return df
 
 
-# ============= 2. STANDARDSCALER NORMALIZATION =============
+# 2. STANDARDSCALER NORMALIZATION
 
 def prepare_normalized_data(df: pd.DataFrame, target_col: str = 'accessibility_score'):
     """
@@ -144,7 +144,7 @@ def prepare_normalized_data(df: pd.DataFrame, target_col: str = 'accessibility_s
     return X_scaled, y_scaled.flatten(), feature_scaler, target_scaler, feature_cols
 
 
-# ============= 3. INCREASED SEQUENCE LENGTH DATASET =============
+# 3. INCREASED SEQUENCE LENGTH DATASET
 
 class EnhancedTimeSeriesDataset(Dataset):
     """
@@ -194,7 +194,7 @@ def __getitem__(self, idx):
         )
 
 
-# ============= 4. TRANSFORMER WITH REGULARIZATION =============
+# 4. TRANSFORMER WITH REGULARIZATION
 
 class RegularizedTransformer(nn.Module):
     """
@@ -257,7 +257,7 @@ def forward(self, x):
         return self.output_head(x)
 
 
-# ============= 5. XGBOOST BASELINE =============
+# 5. XGBOOST BASELINE
 
 def train_xgboost_baseline(X_train, y_train, X_val, y_val, 
                            target_scaler, pred_len: int = 3):
@@ -321,7 +321,7 @@ def train_xgboost_baseline(X_train, y_train, X_val, y_val,
     return models, {'mae': mae, 'rmse': rmse, 'r2': r2}
 
 
-# ============= 6. TRAINING WITH REGULARIZATION =============
+# 6. TRAINING WITH REGULARIZATION
 
 def train_transformer(model, train_loader, val_loader, 
                       target_scaler,