fifa19_streamlit

app.py

@@ -40,70 +40,71 @@ df3, predictors_scaled, predictors_df, train_predictors_val, fifa, df3scaled, xb
 predscale_target = predictors_scaled.columns.tolist()
 
 def player_sim_team(team, position, NUM_RECOM, AGE_upper_bound):
-# part 1(recommendation)
-    target_cols = predscale_target
-
-
-# team stats
-    team_stats = df3scaled.query('position_group == @position and Club == @team').head(3)[target_cols].mean(axis=0)
-    team_stats_np = team_stats.values
-
-# player stats by each position
-    ply_stats = df3scaled.query('position_group == @position and Club != @team and Age1 <= @AGE_upper_bound')[
-    ['ID'] + target_cols]
-    ply_stats_np = ply_stats[target_cols].values
-    X = np.vstack((team_stats_np, ply_stats_np))
-
-## KNN
-    nbrs = NearestNeighbors(n_neighbors=NUM_RECOM + 1, algorithm='auto').fit(X)
-    dist, rank = nbrs.kneighbors(X)
-
-
-    global indice
-    global predicted_players_name
-    global predicted_players_value
-    global predictions
-
-    indice = ply_stats.iloc[rank[0, 1:]].index.tolist()
-    predicted_players_name=df3['Name'].loc[indice,].tolist()
-    predicted_players_value=fifa['Value'].loc[indice,].tolist()
-    display_df1 = predictors_scaled.loc[indice,]
-    playrpredictorss = predictors_df.loc[indice,]
-    display_df2 = df3.loc[indice,]
-    display_df = fifa.loc[indice,]
-
-
-#part 2(prediction)
-    predictors_anomaly_processed=playrpredictorss[playrpredictorss.index.isin(list(display_df2['ID']))].copy()
-    predictors_anomaly_processed['Forward_Skill'] = predictors_anomaly_processed.loc[:,['LS',  'ST', 'RS',  'LW', 'LF', 'CF', 'RF', 'RW']].mean(axis=1)
-
-    predictors_anomaly_processed['Midfield_Skill'] = predictors_anomaly_processed.loc[:,['LAM','CAM','RAM', 'LM', 'LCM', 'CM' ,'RCM', 'RM','LDM', 'CDM', 'RDM']].mean(axis=1)
-
-    predictors_anomaly_processed['Defence_Skill'] = predictors_anomaly_processed.loc[:,['LWB','RWB', 'LB','LCB','CB','RCB','RB']].mean(axis=1)
-
-    predictors_anomaly_processed = predictors_anomaly_processed.drop(['LS',  'ST', 'RS',  'LW', 'LF', 'CF', 'RF', 'RW',
-          'LAM','CAM','RAM', 'LM', 'LCM', 'CM' ,'RCM', 'RM','LDM', 'CDM', 'RDM',
-          'LWB','RWB', 'LB','LCB','CB','RCB','RB'], axis = 1)
-
-    predictors_anomaly_processed=predictors_anomaly_processed.drop(predictors_anomaly_processed.iloc[:,predictors_anomaly_processed.columns.get_loc('Position_CAM'):predictors_anomaly_processed.columns.get_loc('Position_ST')+1], axis=1)
-
-    predictors_anomaly_processed=predictors_anomaly_processed[train_predictors_val.columns]
-    predictors_anomaly_processed[['International Reputation','Real Face']]=predictors_anomaly_processed[['International Reputation','Real Face']].astype('category')
-
-    scaler = StandardScaler()
-    predictors_anomaly_processed[predictors_anomaly_processed.select_dtypes(include=['float64','float32','int64','int32'], exclude=['category']).columns] = scaler.fit_transform(predictors_anomaly_processed.select_dtypes(include=['float64','float32','int64','int32'], exclude=['category']))
-    predictors_anomaly_processed[predictors_anomaly_processed.select_dtypes(include='category').columns]=predictors_anomaly_processed[predictors_anomaly_processed.select_dtypes(include='category').columns].astype('int')
-
-
-    predictions = abs(xbr.predict(predictors_anomaly_processed))
-    predictions = predictions.astype('int64')
-
-    result=final_pred(NUM_RECOM,predictions,predicted_players_value,predicted_players_name)
-    return result
-
-
-
+    try:
+        # part 1(recommendation)
+        target_cols = predscale_target
+
+        # team stats
+        team_stats = df3scaled.query('position_group == @position and Club == @team').head(3)[target_cols].mean(axis=0)
+        team_stats_np = team_stats.values
+
+        # player stats by each position
+        ply_stats = df3scaled.query('position_group == @position and Club != @team and Age1 <= @AGE_upper_bound')[
+        ['ID'] + target_cols]
+        ply_stats_np = ply_stats[target_cols].values
+        X = np.vstack((team_stats_np, ply_stats_np))
+
+        ## KNN
+        nbrs = NearestNeighbors(n_neighbors=NUM_RECOM + 1, algorithm='auto').fit(X)
+        dist, rank = nbrs.kneighbors(X)
+
+        indice = ply_stats.iloc[rank[0, 1:]].index.tolist()
+        predicted_players_name=df3['Name'].loc[indice,].tolist()
+        predicted_players_value=fifa['Value'].loc[indice,].tolist()
+        display_df1 = predictors_scaled.loc[indice,]
+        playrpredictorss = predictors_df.loc[indice,]
+        display_df2 = df3.loc[indice,]
+        display_df = fifa.loc[indice,]
+
+        try:
+            #part 2(prediction)
+            predictors_anomaly_processed=playrpredictorss[playrpredictorss.index.isin(list(display_df2['ID']))].copy()
+            predictors_anomaly_processed['Forward_Skill'] = predictors_anomaly_processed.loc[:,['LS','ST','RS','LW','LF','CF','RF','RW']].mean(axis=1)
+            predictors_anomaly_processed['Midfield_Skill'] = predictors_anomaly_processed.loc[:,['LAM','CAM','RAM','LM','LCM','CM','RCM','RM','LDM','CDM','RDM']].mean(axis=1)
+            predictors_anomaly_processed['Defence_Skill'] = predictors_anomaly_processed.loc[:,['LWB','RWB','LB','LCB','CB','RCB','RB']].mean(axis=1)
+
+            predictors_anomaly_processed = predictors_anomaly_processed.drop(['LS','ST','RS','LW','LF','CF','RF','RW',
+                'LAM','CAM','RAM','LM','LCM','CM','RCM','RM','LDM','CDM','RDM',
+                'LWB','RWB','LB','LCB','CB','RCB','RB'], axis=1)
+
+            predictors_anomaly_processed=predictors_anomaly_processed.drop(predictors_anomaly_processed.iloc[:,predictors_anomaly_processed.columns.get_loc('Position_CAM'):predictors_anomaly_processed.columns.get_loc('Position_ST')+1], axis=1)
+
+            predictors_anomaly_processed=predictors_anomaly_processed[train_predictors_val.columns]
+            predictors_anomaly_processed[['International Reputation','Real Face']]=predictors_anomaly_processed[['International Reputation','Real Face']].astype('category')
+
+            scaler = StandardScaler()
+            predictors_anomaly_processed[predictors_anomaly_processed.select_dtypes(include=['float64','float32','int64','int32'], exclude=['category']).columns] = scaler.fit_transform(predictors_anomaly_processed.select_dtypes(include=['float64','float32','int64','int32'], exclude=['category']))
+            predictors_anomaly_processed[predictors_anomaly_processed.select_dtypes(include='category').columns]=predictors_anomaly_processed[predictors_anomaly_processed.select_dtypes(include='category').columns].astype('int')
+
+            try:
+                predictions = abs(xbr.predict(predictors_anomaly_processed))
+                predictions = predictions.astype('int64')
+            except AttributeError:
+                st.warning("Using fallback prediction method due to model version mismatch")
+                # Fallback to using value directly if prediction fails
+                predictions = predicted_players_value
+
+        except Exception as e:
+            st.error(f"Error in prediction part: {str(e)}")
+            # Fallback to using original values if prediction fails
+            predictions = predicted_players_value
+
+        result = final_pred(NUM_RECOM, predictions, predicted_players_value, predicted_players_name)
+        return result
 
+    except Exception as e:
+        st.error(f"Error in recommendation part: {str(e)}")
+        return []
 
 def final_pred(num_of_players,b=[],c=[],d=[]):
 

requirements.txt

@@ -3,4 +3,4 @@ numpy==1.23.5
 pandas==1.5.3
 scikit-learn==1.2.2
 pickle5
-xgboost==1.7.3 
+xgboost==1.5.2