Update app.py

app.py

@@ -6,8 +6,7 @@ import seaborn as sns
 from datetime import datetime
 from sklearn.metrics import confusion_matrix, precision_score, recall_score
 
-# Sample data preparation (in a real scenario, you would load your data)
-# Converting your sample data to a DataFrame
+# Sample data preparation
 data = {
     'transaction_amount': [2500, 799, 9338, 11749, 8999, 1500, 3000, 4000, 300, 5000, 24990],
     'transaction_date': ['01-11-2024 16:08', '01-11-2024 16:15', '02-11-2024 14:43', '03-11-2024 11:14', 
@@ -27,11 +26,8 @@ data = {
 
 df = pd.DataFrame(data)
 
-# Convert date strings to datetime objects
 df['transaction_date'] = pd.to_datetime(df['transaction_date'], format='%d-%m-%Y %H:%M')
 
-# Add simulated predicted fraud and reported fraud columns
-# In a real scenario, these would come from your model and reports
 np.random.seed(42)
 df['is_fraud_predicted'] = np.random.choice([0, 1], size=len(df), p=[0.3, 0.7])
 df['is_fraud_reported'] = np.random.choice([0, 1], size=len(df), p=[0.4, 0.6])
@@ -39,11 +35,9 @@ df['is_fraud_reported'] = np.random.choice([0, 1], size=len(df), p=[0.4, 0.6])
 def filter_data(start_date, end_date, payer_id, payee_id, transaction_id):
     filtered_df = df.copy()
     
-    # Convert string dates to datetime for comparison
     start_date = pd.to_datetime(start_date)
     end_date = pd.to_datetime(end_date)
     
-    # Apply filters
     filtered_df = filtered_df[(filtered_df['transaction_date'] >= start_date) & 
                              (filtered_df['transaction_date'] <= end_date)]
     
@@ -77,17 +71,14 @@ def create_comparison_chart(dimension, filtered_df):
     else:
         return plt.figure()
     
-    # Group by the selected dimension and count predicted and reported frauds
     predicted = filtered_df.groupby(group_col)['is_fraud_predicted'].sum()
     reported = filtered_df.groupby(group_col)['is_fraud_reported'].sum()
     
-    # Create a DataFrame for plotting
     plot_df = pd.DataFrame({
         'Predicted Fraud': predicted,
         'Reported Fraud': reported
     })
     
-    # Plot
     plot_df.plot(kind='bar', figsize=(10, 6))
     plt.title(f'Fraud Comparison by {dimension}')
     plt.ylabel('Count')
@@ -102,7 +93,6 @@ def create_time_series(filtered_df, granularity):
     
     plt.figure(figsize=(12, 6))
     
-    # Set the time grouping based on granularity
     if granularity == 'Day':
         time_group = filtered_df['transaction_date'].dt.date
     elif granularity == 'Hour':
@@ -112,11 +102,9 @@ def create_time_series(filtered_df, granularity):
     else:
         return plt.figure()
     
-    # Group by time and count predicted and reported frauds
     predicted = filtered_df.groupby(time_group)['is_fraud_predicted'].sum()
     reported = filtered_df.groupby(time_group)['is_fraud_reported'].sum()
     
-    # Plot
     plt.plot(predicted.index, predicted.values, 'b-', label='Predicted Fraud')
     plt.plot(reported.index, reported.values, 'r-', label='Reported Fraud')
     plt.title('Fraud Trend Over Time')
@@ -132,14 +120,11 @@ def calculate_metrics(filtered_df):
     if filtered_df.empty:
         return None, 0, 0
     
-    # Calculate confusion matrix
     cm = confusion_matrix(filtered_df['is_fraud'], filtered_df['is_fraud_predicted'])
     
-    # Calculate precision and recall
     precision = precision_score(filtered_df['is_fraud'], filtered_df['is_fraud_predicted'], zero_division=0)
     recall = recall_score(filtered_df['is_fraud'], filtered_df['is_fraud_predicted'], zero_division=0)
     
-    # Create confusion matrix plot
     plt.figure(figsize=(6, 5))
     sns.heatmap(cm, annot=True, fmt='d', cmap='Blues', 
                 xticklabels=['Not Fraud', 'Fraud'],
@@ -151,19 +136,14 @@ def calculate_metrics(filtered_df):
     return plt, precision, recall
 
 def update_interface(start_date, end_date, payer_id, payee_id, transaction_id, dimension, time_granularity):
-    # Filter data based on inputs
     filtered_df = filter_data(start_date, end_date, payer_id, payee_id, transaction_id)
     
-    # Create comparison chart
     comparison_chart = create_comparison_chart(dimension, filtered_df)
     
-    # Create time series chart
     time_series = create_time_series(filtered_df, time_granularity)
     
-    # Calculate evaluation metrics
     confusion_matrix_plot, precision, recall = calculate_metrics(filtered_df)
     
-    # Format the filtered dataframe for display
     display_df = filtered_df.copy()
     display_df['transaction_date'] = display_df['transaction_date'].dt.strftime('%Y-%m-%d %H:%M')
     
@@ -174,7 +154,6 @@ def update_interface(start_date, end_date, payer_id, payee_id, transaction_id, d
             f"Precision: {precision:.4f}", 
             f"Recall: {recall:.4f}")
 
-# Define the Gradio interface
 with gr.Blocks() as demo:
     gr.Markdown("# Fraud Transaction Analysis Dashboard")
     
@@ -233,6 +212,5 @@ with gr.Blocks() as demo:
         outputs=[data_table, comparison_plot, time_series_plot, confusion_matrix_plot, precision_text, recall_text]
     )
 
-# Launch the app
 if __name__ == "__main__":
     demo.launch()