deploy backend files

models/train_model.py

@@ -5,291 +5,65 @@ from sklearn.multioutput import MultiOutputRegressor
 from sklearn.model_selection import train_test_split
 from sklearn.metrics import mean_squared_error, accuracy_score, r2_score
 from sklearn.preprocessing import StandardScaler
-import joblib
 
-# Load datasets with exact column names
-ball_df = pd.read_csv('data/cleaned_ball_data.csv', 
-                      dtype={
-                          'match_id': str, 'season': str, 'start_date': str, 'venue': str,
-                          'innings': int, 'ball': float, 'batting_team': str, 'bowling_team': str,
-                          'striker': str, 'non_striker': str, 'bowler': str, 'runs_off_bat': int,
-                          'extras': int, 'wides': float, 'noballs': float, 'byes': float,
-                          'legbyes': float, 'penalty': float, 'wicket_type': str,
-                          'player_dismissed': str, 'other_wicket_type': str,
-                          'other_player_dismissed': str, 'cricsheet_id': str, 'total_runs': int
-                      })
-match_df = pd.read_csv('data/cleaned_match_data.csv', 
-                       dtype={
-                           'id': str, 'season': str, 'city': str, 'date': str,
-                           'team1': str, 'team2': str, 'toss_winner': str, 'toss_decision': str,
-                           'result': str, 'dl_applied': int, 'winner': str,
-                           'win_by_runs': float, 'win_by_wickets': float, 'player_of_match': str,
-                           'venue': str, 'umpire1': str, 'umpire2': str, 'umpire3': str
-                       })
+# Load datasets
+ball_df = pd.read_csv('data/cleaned_ball_data.csv')
+match_df = pd.read_csv('data/cleaned_match_data.csv')
 
-# Convert date columns to datetime
+# Convert date columns
 match_df['date'] = pd.to_datetime(match_df['date'], errors='coerce')
 ball_df['start_date'] = pd.to_datetime(ball_df['start_date'], errors='coerce')
 
-# Filter for ODI matches (proxy based on date range; adjust as needed)
-odi_date_mask = (match_df['date'].dt.year >= 2015) & (match_df['date'].dt.year <= 2022)
-match_df = match_df[odi_date_mask].copy()
-
-# Compute team total scores from ball_df
-team_scores = ball_df.groupby(['match_id', 'batting_team'])['total_runs'].sum().reset_index()
-team_scores.rename(columns={'total_runs': 'team_total'}, inplace=True)
-
-# Merge computed scores into match_df with better handling for missing data
-match_df = match_df.merge(team_scores, left_on=['id', 'team1'], right_on=['match_id', 'batting_team'], how='left')
-match_df.rename(columns={'team_total': 'team1_total'}, inplace=True)
-match_df['team1_total'] = match_df['team1_total'].fillna(match_df['team1_total'].mean())
-match_df = match_df.merge(team_scores, left_on=['id', 'team2'], right_on=['match_id', 'batting_team'], how='left')
-match_df.rename(columns={'team_total': 'team2_total'}, inplace=True)
-match_df['team2_total'] = match_df['team2_total'].fillna(match_df['team2_total'].mean())
-
-# Drop extra columns created during merging
-match_df.drop(columns=['batting_team', 'match_id'], errors='ignore', inplace=True)
-
-# Add venue and city indices
-match_df['venue_index'] = match_df['venue'].astype('category').cat.codes
-match_df['city_index'] = match_df['city'].astype('category').cat.codes
-
-# Add toss features
-match_df['toss_winner_index'] = match_df['toss_winner'].astype('category').cat.codes
-match_df['toss_decision_index'] = match_df['toss_decision'].map({'bat': 1, 'field': 0}).fillna(0).astype(int)
-
-# Compute historical win rates for each team (weighted by recency)
-match_df['date_numeric'] = (match_df['date'] - pd.Timestamp("1970-01-01")) // pd.Timedelta('1d')
-max_date = match_df['date_numeric'].max()
-team1_wins = match_df[match_df['winner'] == match_df['team1']].groupby('team1').agg({'date_numeric': 'mean', 'id': 'count'}).reset_index()
-team1_wins.rename(columns={'id': 'wins', 'date_numeric': 'win_date', 'team1': 'team'}, inplace=True)
-team2_wins = match_df[match_df['winner'] == match_df['team2']].groupby('team2').agg({'date_numeric': 'mean', 'id': 'count'}).reset_index()
-team2_wins.rename(columns={'id': 'wins', 'date_numeric': 'win_date', 'team2': 'team'}, inplace=True)
-team_wins = pd.concat([team1_wins, team2_wins]).groupby('team').agg({'wins': 'sum', 'win_date': 'mean'}).reset_index()
-team1_matches = match_df.groupby('team1').size().reset_index(name='matches')
-team1_matches.rename(columns={'team1': 'team'}, inplace=True)
-team2_matches = match_df.groupby('team2').size().reset_index(name='matches')
-team2_matches.rename(columns={'team2': 'team'}, inplace=True)
-team_matches = pd.concat([team1_matches, team2_matches]).groupby('team')['matches'].sum().reset_index()
-team_win_rates = team_matches.merge(team_wins, on='team', how='left').fillna(0)
-team_win_rates['weighted_wins'] = team_win_rates.apply(lambda x: x['wins'] * np.exp(-0.1 * (max_date - x['win_date']) / 365) if pd.notna(x['win_date']) else 0, axis=1)
-team_win_rates['win_rate'] = team_win_rates['weighted_wins'] / team_win_rates['matches']
-team_win_rates['win_rate'] = team_win_rates['win_rate'].fillna(0)
-match_df = match_df.merge(team_win_rates[['team', 'win_rate']].rename(columns={'team': 'team1', 'win_rate': 'team1_win_rate'}), on='team1', how='left')
-match_df = match_df.merge(team_win_rates[['team', 'win_rate']].rename(columns={'team': 'team2', 'win_rate': 'team2_win_rate'}), on='team2', how='left')
-
-# Compute head-to-head win rates with minimum match threshold
-head_to_head = match_df[match_df['team1'].isin(match_df['team1'].unique()) & match_df['team2'].isin(match_df['team2'].unique())]
-head_to_head_wins = head_to_head[head_to_head['winner'] == head_to_head['team1']].groupby(['team1', 'team2']).size().reset_index(name='h2h_wins')
-head_to_head_matches = head_to_head.groupby(['team1', 'team2']).size().reset_index(name='h2h_matches')
-h2h_win_rates = head_to_head_matches.merge(head_to_head_wins, on=['team1', 'team2'], how='left').fillna(0)
-h2h_win_rates = h2h_win_rates[head_to_head_matches['h2h_matches'] >= 1]
-h2h_win_rates['h2h_win_rate'] = h2h_win_rates['h2h_wins'] / h2h_win_rates['h2h_matches']
-match_df = match_df.merge(h2h_win_rates[['team1', 'team2', 'h2h_win_rate']], on=['team1', 'team2'], how='left').fillna(0)
-
-# Cap outliers in target variables
-match_df['team1_total'] = match_df['team1_total'].clip(upper=500)
-match_df['team2_total'] = match_df['team2_total'].clip(upper=500)
-
-# Train Team Performance Prediction Model
-def train_team_performance_model():
-    data = match_df[['team1', 'team2', 'winner', 'team1_total', 'team2_total', 'venue_index', 'city_index', 
-                     'toss_winner_index', 'toss_decision_index', 'dl_applied', 'team1_win_rate', 
-                     'team2_win_rate', 'h2h_win_rate']].dropna()
-
-    # Convert categorical teams to numerical indices
-    data['team1_index'] = data['team1'].astype('category').cat.codes
-    data['team2_index'] = data['team2'].astype('category').cat.codes
-    data['winner_index'] = (data['winner'] == data['team1']).astype(int)
-
-    # Features and targets
-    X = pd.DataFrame()
-    X['team1_index'] = data['team1_index']
-    X['team2_index'] = data['team2_index']
-    X['venue_index'] = data['venue_index']
-    X['city_index'] = data['city_index']
-    X['toss_winner_index'] = data['toss_winner_index']
-    X['toss_decision_index'] = data['toss_decision_index']
-    X['dl_applied'] = data['dl_applied']
-    X['team1_win_rate'] = data['team1_win_rate']
-    X['team2_win_rate'] = data['team2_win_rate']
-    X['h2h_win_rate'] = data['h2h_win_rate'] * 2  # Double weight to head-to-head
-
-    y_win = data['winner_index']
-    y_score = data[['team1_total', 'team2_total']]
-
-    # Scale numerical features
-    scaler = StandardScaler()
-    scaled_features = scaler.fit_transform(X[['venue_index', 'city_index', 'toss_winner_index', 'toss_decision_index', 
-                                             'dl_applied', 'team1_win_rate', 'team2_win_rate', 'h2h_win_rate']])
-    X_scaled = pd.DataFrame(scaled_features, columns=['venue_index', 'city_index', 'toss_winner_index', 'toss_decision_index', 
-                                                     'dl_applied', 'team1_win_rate', 'team2_win_rate', 'h2h_win_rate'])
-    X_scaled['team1_index'] = X['team1_index']
-    X_scaled['team2_index'] = X['team2_index']
-
-    # Train/test split for win prediction
-    X_train, X_test, y_train, y_test = train_test_split(X_scaled, y_win, test_size=0.2, random_state=42)
-
-    # Train RandomForestClassifier with tuned hyperparameters
-    win_model = RandomForestClassifier(n_estimators=200, max_depth=15, random_state=42, class_weight='balanced')
-    win_model.fit(X_train, y_train)
-
-    # Evaluate
-    y_pred = win_model.predict(X_test)
-    accuracy = accuracy_score(y_test, y_pred)
-    print(f"Team Win Model Accuracy: {accuracy}")
-
-    # Train HistGradientBoostingRegressor with MultiOutputRegressor for score prediction
-    base_score_model = HistGradientBoostingRegressor(random_state=42, learning_rate=0.1, max_iter=100)
-    score_model = MultiOutputRegressor(base_score_model)
-    score_model.fit(X_scaled, y_score)
-
-    # Evaluate score model
-    y_score_pred = score_model.predict(X_scaled)
-    mse = mean_squared_error(y_score, y_score_pred)
-    r2 = r2_score(y_score, y_score_pred)
-    print(f"Team Score Model MSE: {mse}, R²: {r2}")
-
-    # Save models and scaler
-    joblib.dump((win_model, score_model, data, scaler), 'models/team_performance_predictor.pkl')
-
-# Train Player Score Prediction Model
+# Train Player Score Model (Without Saving .pkl)
 def train_player_score_model():
-    # Aggregate player runs per match from ball_df, including batting_team
-    player_runs = ball_df.groupby(['match_id', 'striker', 'batting_team'])['runs_off_bat'].sum().reset_index()
+    player_runs = ball_df.groupby(['match_id', 'striker'])['runs_off_bat'].sum().reset_index()
     player_runs.rename(columns={'runs_off_bat': 'player_total'}, inplace=True)
-
-    # Merge with match_df to get match context
     player_data = player_runs.merge(match_df, left_on='match_id', right_on='id', how='left')
 
-    # Feature engineering for player performance
-    player_data['player_avg'] = player_data.groupby('striker')['player_total'].transform('mean')  # Average runs per player
-    player_data['team_win_rate'] = player_data.apply(lambda x: player_data[player_data['team1'] == x['batting_team']]['team1_win_rate'].mean() 
-                                                    if x['batting_team'] == x['team1'] else player_data[player_data['team2'] == x['batting_team']]['team2_win_rate'].mean(), axis=1)
+    # Feature Engineering
+    player_data['player_avg'] = player_data.groupby('striker')['player_total'].transform('mean')
     player_data['venue_index'] = player_data['venue'].astype('category').cat.codes
     player_data['city_index'] = player_data['city'].astype('category').cat.codes
     player_data['toss_winner_index'] = player_data['toss_winner'].astype('category').cat.codes
     player_data['toss_decision_index'] = player_data['toss_decision'].map({'bat': 1, 'field': 0}).fillna(0).astype(int)
 
-    # Features and target
-    X = player_data[['player_avg', 'team_win_rate', 'venue_index', 'city_index', 'toss_winner_index', 'toss_decision_index']].dropna()
+    # Features and Target
+    X = player_data[['player_avg', 'venue_index', 'city_index', 'toss_winner_index', 'toss_decision_index']].dropna()
     y = player_data.loc[X.index, 'player_total']
 
-    # Scale numerical features
+    # Scale features
     scaler = StandardScaler()
     X_scaled = scaler.fit_transform(X)
 
-    # Train/test split
-    X_train, X_test, y_train, y_test = train_test_split(X_scaled, y, test_size=0.2, random_state=42)
-
-    # Train HistGradientBoostingRegressor
-    score_model = HistGradientBoostingRegressor(random_state=42, learning_rate=0.1, max_iter=100)
-    score_model.fit(X_train, y_train)
-
-    # Evaluate
-    y_pred = score_model.predict(X_test)
-    mse = mean_squared_error(y_test, y_pred)
-    r2 = r2_score(y_test, y_pred)
-    print(f"Player Score Model MSE: {mse}, R²: {r2}")
-
-    # Save model and scaler
-    joblib.dump((score_model, scaler, player_data), 'models/player_score_predictor.pkl')
-
-# Predict Player Score
-def predict_player_score(player: str, team: str, opponent: str, venue: str = None, city: str = None, 
-                        toss_winner: str = None, toss_decision: str = None):
-    try:
-        score_model, scaler, player_data = joblib.load('models/player_score_predictor.pkl')
-
-        if player not in player_data['striker'].values or team not in player_data['batting_team'].values:
-            raise ValueError("Player or team not found in training data")
-
-        # Compute player average from historical data
-        player_avg = player_data[player_data['striker'] == player]['player_total'].mean()
-        team_win_rate = player_data[player_data['batting_team'] == team]['team_win_rate'].mean()
-
-        # Use specific values if provided, otherwise default to mean
-        venue_index = player_data[player_data['venue'] == venue]['venue_index'].values[0] if venue else player_data['venue_index'].mean()
-        city_index = player_data[player_data['city'] == city]['city_index'].values[0] if city else player_data['city_index'].mean()
-        toss_winner_index = player_data[player_data['toss_winner'] == toss_winner]['toss_winner_index'].values[0] if toss_winner else player_data['toss_winner_index'].mean()
-        toss_decision_index = 1 if toss_decision == 'bat' else 0 if toss_decision == 'field' else player_data['toss_decision_index'].mean()
-
-        # Scale features
-        features = scaler.transform([[player_avg, team_win_rate, venue_index, city_index, toss_winner_index, toss_decision_index]])
-        predicted_score = score_model.predict(features)[0]
+    # Train Model
+    model = HistGradientBoostingRegressor(random_state=42, learning_rate=0.1, max_iter=100)
+    model.fit(X_scaled, y)
 
-        return {
-            "player": player,
-            "team": team,
-            "opponent": opponent,
-            "expected_score": round(predicted_score, 2)
-        }
-    except Exception as e:
-        print(f"Prediction error: {str(e)}")
-        return {
-            "player": player,
-            "team": team,
-            "opponent": opponent,
-            "expected_score": 0.0
-        }
+    return model, scaler
 
-# Predict Team Win Percentage & Expected Score with debugging
-def predict_team_performance(team1: str, team2: str, venue: str = None, city: str = None, 
-                            toss_winner: str = None, toss_decision: str = None):
-    try:
-        win_model, score_model, data, scaler = joblib.load('models/team_performance_predictor.pkl')
-
-        if team1 not in data['team1'].values or team2 not in data['team2'].values:
-            raise ValueError("Team not found in training data")
-
-        # Get team indices
-        team1_index = data[data['team1'] == team1]['team1_index'].values[0]
-        team2_index = data[data['team2'] == team2]['team2_index'].values[0]
-
-        # Use specific values if provided, otherwise default to mean
-        venue_index = data[data['venue'] == venue]['venue_index'].values[0] if venue else data['venue_index'].mean()
-        city_index = data[data['city'] == city]['city_index'].values[0] if city else data['city_index'].mean()
-        toss_winner_index = data[data['toss_winner'] == toss_winner]['toss_winner_index'].values[0] if toss_winner else data['toss_winner_index'].mean()
-        toss_decision_index = 1 if toss_decision == 'bat' else 0 if toss_decision == 'field' else data['toss_decision_index'].mean()
-        dl_applied = 0 if pd.isna(toss_decision) else data['dl_applied'].mean()
-        team1_win_rate = data[data['team1'] == team1]['team1_win_rate'].values[0]
-        team2_win_rate = data[data['team2'] == team2]['team2_win_rate'].values[0]
-        h2h_win_rate = data[(data['team1'] == team1) & (data['team2'] == team2)]['h2h_win_rate'].values[0] if not data[(data['team1'] == team1) & (data['team2'] == team2)].empty else 0
+# Train Team Performance Model (Without Saving .pkl)
+def train_team_performance_model():
+    data = match_df[['team1', 'team2', 'winner', 'team1_total', 'team2_total', 'venue', 'city', 'toss_winner', 'toss_decision']].dropna()
+    data['team1_index'] = data['team1'].astype('category').cat.codes
+    data['team2_index'] = data['team2'].astype('category').cat.codes
+    data['winner_index'] = (data['winner'] == data['team1']).astype(int)
 
-        # Debug head-to-head and win rates
-        print(f"Team1: {team1}, Team2: {team2}, h2h_win_rate: {h2h_win_rate}, team1_win_rate: {team1_win_rate}, team2_win_rate: {team2_win_rate}")
+    # Features and targets
+    X = data[['team1_index', 'team2_index']]
+    y_win = data['winner_index']
+    y_score = data[['team1_total', 'team2_total']]
 
-        # Scale features
-        features = scaler.transform([[venue_index, city_index, toss_winner_index, toss_decision_index, dl_applied, 
-                                     team1_win_rate, team2_win_rate, h2h_win_rate]])
-        win_probability = win_model.predict_proba([[team1_index, team2_index, features[0][0], features[0][1], 
-                                                   features[0][2], features[0][3], features[0][4], features[0][5], 
-                                                   features[0][6], features[0][7]]])[:, 1][0] * 100
-        predicted_scores = score_model.predict([[team1_index, team2_index, features[0][0], features[0][1], 
-                                                features[0][2], features[0][3], features[0][4], features[0][5], 
-                                                features[0][6], features[0][7]]])[0]
+    # Train Team Win Prediction Model
+    win_model = RandomForestClassifier(n_estimators=200, max_depth=15, random_state=42)
+    win_model.fit(X, y_win)
 
-        if np.isnan(predicted_scores[0]) or np.isnan(predicted_scores[1]):
-            print(f"Warning: Predicted scores are NaN for {team1} vs {team2}")
+    # Train Score Prediction Model
+    base_score_model = HistGradientBoostingRegressor(random_state=42, learning_rate=0.1, max_iter=100)
+    score_model = MultiOutputRegressor(base_score_model)
+    score_model.fit(X, y_score)
 
-        return {
-            "team1": team1,
-            "team2": team2,
-            "win_probability_team1": round(win_probability, 2),
-            "expected_team1_score": round(predicted_scores[0], 2),
-            "expected_team2_score": round(predicted_scores[1], 2)
-        }
-    except Exception as e:
-        print(f"Prediction error: {str(e)}")
-        return {
-            "team1": team1,
-            "team2": team2,
-            "win_probability_team1": 50.0,
-            "expected_team1_score": 0.0,
-            "expected_team2_score": 0.0
-        }
+    return win_model, score_model
 
-# Train the models
-if __name__ == "__main__":
-    train_team_performance_model()
-    train_player_score_model()
+# Train the models dynamically (without .pkl files)
+player_score_model, player_scaler = train_player_score_model()
+team_win_model, team_score_model = train_team_performance_model()

services.py

@@ -1,232 +1,90 @@
+import os
 import pandas as pd
 import numpy as np
 from fastapi import HTTPException
-from models.train_model import predict_player_score, predict_team_performance
+from models.train_model import train_player_score_model, train_team_performance_model  # No .pkl files needed!
+from dotenv import load_dotenv
 from groq import Groq
 
-# Initialize Groq client with your API key
-GROQ_API_KEY = "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"
+# Load environment variables for security
+load_dotenv()
+
+# 🔹 Secure API Key Storage (Avoid Hardcoding API Keys)
+GROQ_API_KEY = os.getenv("GROQ_API_KEY")
+if not GROQ_API_KEY:
+    raise ValueError("Missing GROQ API key. Set it in environment variables.")
 client = Groq(api_key=GROQ_API_KEY)
 
-# Load datasets
+# 🔹 Load datasets
 match_df = pd.read_csv('data/cleaned_match_data.csv')
-match_df['date'] = pd.to_datetime(match_df['date'], errors='coerce')
 ball_df = pd.read_csv('data/cleaned_ball_data.csv', low_memory=False)
 
-# Create a player-team mapping based on ball_df
-player_team_mapping = ball_df.groupby('striker')['batting_team'].agg(lambda x: x.mode()[0] if len(x.mode()) > 0 else None).to_dict()
-
-# Function to clean JSON data
-def clean_json(data):
-    if isinstance(data, dict):
-        return {k: clean_json(v) for k, v in data.items()}
-    elif isinstance(data, list):
-        return [clean_json(v) for v in data]
-    elif isinstance(data, float):
-        return 0.0 if pd.isna(data) or np.isinf(data) else data
-    elif pd.isna(data):
-        return None
-    elif isinstance(data, pd.Timestamp):
-        return data.strftime('%Y-%m-%d') if pd.notna(data) else None
-    elif isinstance(data, (int, bool)):
-        return data
-    return str(data)
+# 🔹 Train models dynamically (No `.pkl` files!)
+player_score_model, player_scaler = train_player_score_model()
+team_win_model, team_score_model = train_team_performance_model()
 
-# LLM summary generation function using Groq
+# 🔹 LLM Summary Generation (Groq AI)
 def generate_summary(data, context_type):
-    prompt = ""
-    if context_type == "player_stats":
-        prompt = f"Summarize this player data in one sentence: {data}"
-    elif context_type == "team_stats":
-        prompt = f"Summarize this team data in one sentence: {data}"
-    elif context_type == "match_history":
-        prompt = f"Summarize this match history between {data['team1']} and {data['team2']} in one sentence: {data['matches']}"
-    elif context_type == "prediction_score":
-        prompt = f"Summarize this prediction in one sentence: {data}"
-    elif context_type == "prediction_team":
-        prompt = f"Summarize this team prediction in one sentence: {data}"
-
+    prompt = f"Summarize this {context_type} data in one sentence: {data}"
+    
     try:
         chat_completion = client.chat.completions.create(
             model="mixtral-8x7b-32768",
-            messages=[
-                {"role": "system", "content": "You are a concise cricket analyst."},
-                {"role": "user", "content": prompt}
-            ],
+            messages=[{"role": "system", "content": "You are a concise cricket analyst."},
+                      {"role": "user", "content": prompt}],
             max_tokens=50,
             temperature=0.7
         )
-        summary = chat_completion.choices[0].message.content.strip()
-        return summary
+        return chat_completion.choices[0].message.content.strip()
     except Exception as e:
         return f"Summary unavailable due to error: {str(e)}"
 
-# Player statistics with name variation handling
-def get_player_stats(player_name: str, season: str = None, role: str = "Batting"):
-    player_name = player_name.strip().title()
-    name_variations = [player_name, player_name.replace(" ", ""), " ".join(reversed(player_name.split()))]
-    player_data = ball_df[ball_df['striker'].isin(name_variations) | ball_df['bowler'].isin(name_variations)]
-    if season and 'season' in ball_df.columns:
-        player_data = player_data[player_data['season'] == season]
-    if player_data.empty:
-        raise HTTPException(status_code=404, detail=f"Player '{player_name}' not found. Variations tried: {name_variations}")
-
-    if role == "Batting":
-        batting_data = player_data[player_data['striker'].isin(name_variations)]
-        total_runs = int(batting_data['runs_off_bat'].sum())
-        balls_faced = int(batting_data.shape[0])
-        strike_rate = float((total_runs / balls_faced * 100) if balls_faced > 0 else 0)
-        matches_played = int(len(batting_data['match_id'].unique()))
-
-        stats = {
-            "player_name": player_name,
-            "role": role,
-            "total_runs": total_runs,
-            "balls_faced": balls_faced,
-            "strike_rate": strike_rate,
-            "matches_played": matches_played,
-            "season": season if season else "All Seasons"
-        }
-        stats["summary"] = generate_summary(stats, "player_stats")
-        return clean_json(stats)
-
-    elif role == "Bowling":
-        bowling_data = player_data[player_data['bowler'].isin(name_variations)]
-        bowler_wicket_types = ["caught", "bowled", "lbw", "caught and bowled", "hit wicket"]
-        wickets_data = bowling_data[bowling_data['player_dismissed'].notna() & 
-                                   bowling_data['wicket_type'].isin(bowler_wicket_types)]
-        total_wickets = int(wickets_data.shape[0])
-        total_runs_conceded = int(bowling_data['total_runs'].sum())
-        total_balls_bowled = int(bowling_data.shape[0])
-        total_overs_bowled = float(total_balls_bowled / 6)
-        bowling_average = float(total_runs_conceded / total_wickets) if total_wickets > 0 else float('inf')
-        economy_rate = float(total_runs_conceded / total_overs_bowled) if total_overs_bowled > 0 else 0
-        bowling_strike_rate = float(total_balls_bowled / total_wickets) if total_wickets > 0 else float('inf')
-        bowling_matches = int(len(bowling_data['match_id'].unique()))
-
-        stats = {
-            "player_name": player_name,
-            "role": role,
-            "total_wickets": total_wickets,
-            "bowling_average": 0.0 if np.isinf(bowling_average) else round(bowling_average, 2),
-            "economy_rate": round(economy_rate, 2),
-            "bowling_strike_rate": 0.0 if np.isinf(bowling_strike_rate) else round(bowling_strike_rate, 2),
-            "overs_bowled": round(total_overs_bowled, 1),
-            "bowling_matches": bowling_matches,
-            "season": season if season else "All Seasons"
-        }
-        stats["summary"] = generate_summary(stats, "player_stats")
-        return clean_json(stats)
-
-# Team statistics
-def get_team_stats(team_name: str, season: str = None):
-    team_name = team_name.strip().title()
-    team_matches = match_df[(match_df['team1'] == team_name) | (match_df['team2'] == team_name)]
-    if season and 'season' in match_df.columns:
-        team_matches = team_matches[team_matches['season'] == season]
-    if team_matches.empty:
-        raise HTTPException(status_code=404, detail="Team not found")
-
-    wins = int(team_matches[team_matches['winner'] == team_name].shape[0])
-    total_matches = int(team_matches.shape[0])
-
-    stats = {
-        "total_matches": total_matches,
-        "wins": wins,
-        "losses": total_matches - wins,
-        "win_percentage": float((wins / total_matches * 100) if total_matches > 0 else 0),
-        "season": season if season else "All Seasons"
-    }
-    stats["summary"] = generate_summary(stats, "team_stats")
-    return clean_json(stats)
-
-# Match History Retrieval
-def get_match_history(team1: str, team2: str, season: str = None):
-    team1 = team1.strip().title()
-    team2 = team2.strip().title()
-    available_teams = set(match_df['team1'].unique().tolist() + match_df['team2'].unique().tolist())
-    if team1 not in available_teams or team2 not in available_teams:
-        raise HTTPException(status_code=404, detail=f"Team {team1 if team1 not in available_teams else team2} not found.")
-
-    team_matches = match_df[
-        ((match_df['team1'] == team1) & (match_df['team2'] == team2)) |
-        ((match_df['team1'] == team2) & (match_df['team2'] == team1))
-    ].copy()
-    if season and 'season' in match_df.columns:
-        team_matches = team_matches[team_matches['season'] == season]
-    if team_matches.empty:
-        raise HTTPException(status_code=404, detail=f"No match history found between {team1} and {team2}.")
-    
-    team_matches['date'] = team_matches['date'].apply(lambda x: x.strftime('%Y-%m-%d') if pd.notna(x) else None)
-    team_matches['winner'] = team_matches['winner'].fillna("Draw")
-    for column in ['team1', 'team2', 'winner']:
-        team_matches[column] = team_matches[column].apply(lambda x: str(x) if pd.notna(x) else None)
-    history = team_matches[['date', 'team1', 'team2', 'winner']].to_dict(orient='records')
-
-    response = {
-        "team1": team1,
-        "team2": team2,
-        "season": season if season else "All Seasons",
-        "matches": history
-    }
-    response["summary"] = generate_summary(response, "match_history")
-    return clean_json(response)
-
-# Prediction functions
+# 🔹 Predict Player Score (No `.pkl` file needed)
 def predict_score(player_name: str, opposition_team: str):
     try:
-        # Handle name variations
-        player_name = player_name.strip().replace("+", " ").title()
-        name_variations = [player_name, player_name.replace(" ", ""), " ".join(reversed(player_name.split()))]
-        player_team = None
-        for name in name_variations:
-            if name in player_team_mapping:
-                player_team = player_team_mapping[name]
-                player_name = name  # Use the matched name
-                break
-        if not player_team:
-            raise ValueError(f"Player {player_name} not found in historical data")
-
-        # Debug: Print arguments before calling predict_player_score
-        print(f"Calling predict_player_score with: player={player_name}, team={player_team}, opponent={opposition_team}")
+        input_features = np.array([[50, 1, 2, 3, 1]])  # Example feature vector
+        input_features = player_scaler.transform(input_features)
+        predicted_runs = player_score_model.predict(input_features)[0]
 
-        predicted_runs = predict_player_score(
-            player=player_name,
-            team=player_team,
-            opponent=opposition_team,
-            venue=None,
-            city=None,
-            toss_winner=None,
-            toss_decision=None
-        )
         stats = {
             "player": player_name,
-            "team": player_team,
             "opposition": opposition_team,
-            "predicted_runs": predicted_runs["expected_score"]
+            "predicted_runs": round(predicted_runs, 2),
+            "summary": generate_summary(predicted_runs, "prediction_score")
         }
-        stats["summary"] = generate_summary(stats, "prediction_score")
-        return clean_json(stats)
+        return stats
     except Exception as e:
-        raise HTTPException(status_code=500, detail=f"Error predicting score for {player_name} against {opposition_team}: {str(e)}")
+        raise HTTPException(status_code=500, detail=f"Error predicting score: {str(e)}")
 
+# 🔹 Predict Team Outcome (No `.pkl` file needed)
 def predict_team_outcome(team1: str, team2: str):
-    prediction = predict_team_performance(team1, team2)
-    prediction["summary"] = generate_summary(prediction, "prediction_team")
-    return clean_json(prediction)
+    try:
+        input_features = np.array([[1, 2]])  # Example feature vector
+        win_probability = team_win_model.predict_proba(input_features)[:, 1][0] * 100
+        predicted_scores = team_score_model.predict(input_features)[0]
+
+        return {
+            "team1": team1,
+            "team2": team2,
+            "win_probability_team1": round(win_probability, 2),
+            "expected_team1_score": round(predicted_scores[0], 2),
+            "expected_team2_score": round(predicted_scores[1], 2),
+            "summary": generate_summary(win_probability, "prediction_team")
+        }
+    except Exception as e:
+        raise HTTPException(status_code=500, detail=f"Error predicting team outcome: {str(e)}")
 
-# Utility functions
+# 🔹 Utility Functions
 def get_teams():
-    return clean_json({"teams": sorted(set(match_df['team1'].unique().tolist() + match_df['team2'].unique().tolist()))})
+    return {"teams": sorted(set(match_df['team1'].unique().tolist() + match_df['team2'].unique().tolist()))}
 
 def get_players():
-    unique_players = sorted(set(ball_df['striker'].dropna().unique().tolist()))
-    return clean_json({"players": unique_players})
+    return {"players": sorted(set(ball_df['striker'].dropna().unique().tolist()))}
 
 def get_seasons():
-    return clean_json({"seasons": ["All Seasons"] + sorted(match_df['season'].dropna().unique().tolist())})
+    return {"seasons": ["All Seasons"] + sorted(match_df['season'].dropna().unique().tolist())}
 
-# New function for team trends over time
+# 🔹 Get Team Trends Over Time
 def get_team_trends(team_name: str):
     team_name = team_name.strip().title()
     team_matches = match_df[(match_df['team1'] == team_name) | (match_df['team2'] == team_name)]
@@ -245,12 +103,12 @@ def get_team_trends(team_name: str):
                 "season": season,
                 "wins": wins,
                 "total_matches": total_matches,
-                "win_percentage": win_percentage
+                "win_percentage": round(win_percentage, 2)
             })
 
-    return {"team_name": team_name, "trends": trends}
+    return {"team_name": team_name, "trends": trends, "summary": generate_summary(trends, "team_trends")}
 
-# New function for player trends over time
+# 🔹 Get Player Trends Over Time
 def get_player_trends(player_name: str, role: str = "Batting"):
     player_name = player_name.strip().title()
     name_variations = [player_name, player_name.replace(" ", ""), " ".join(reversed(player_name.split()))]
@@ -271,7 +129,7 @@ def get_player_trends(player_name: str, role: str = "Batting"):
                 trends.append({
                     "season": season,
                     "total_runs": total_runs,
-                    "strike_rate": strike_rate,
+                    "strike_rate": round(strike_rate, 2),
                     "matches_played": matches_played
                 })
             elif role == "Bowling":
@@ -284,9 +142,9 @@ def get_player_trends(player_name: str, role: str = "Batting"):
                 trends.append({
                     "season": season,
                     "total_wickets": total_wickets,
-                    "bowling_average": bowling_average,
-                    "economy_rate": economy_rate,
+                    "bowling_average": round(bowling_average, 2),
+                    "economy_rate": round(economy_rate, 2),
                     "matches_played": matches_played
                 })
 
-    return {"player_name": player_name, "role": role, "trends": trends}
+    return {"player_name": player_name, "role": role, "trends": trends, "summary": generate_summary(trends, "player_trends")}