Create app.py

app.py

@@ -0,0 +1,117 @@
+import pandas as pd
+import numpy as np
+from sklearn.model_selection import train_test_split
+from sklearn.preprocessing import StandardScaler
+from sklearn.utils import resample
+from sklearn.metrics import accuracy_score, classification_report
+from sklearn.linear_model import LogisticRegression
+from imblearn.over_sampling import SMOTE
+from transformers import pipeline
+import gradio as gr
+from google.colab import files
+
+# Load the creditcard.csv dataset from Google Drive or Colab local file upload
+uploaded = files.upload()  # This will prompt you to upload your file
+df = pd.read_csv('creditcard.csv')
+
+# Display basic information
+print("Columns in the dataset:", df.columns)
+print(df.head())
+
+# Preprocessing: Selecting relevant columns
+# Assuming the dataset has 'Time', 'Amount', and 'Class' columns along with 'V1' to 'V28' features
+time_col = 'Time'
+amount_col = 'Amount'
+class_col = 'Class'
+feature_cols = [col for col in df.columns if col not in [class_col, time_col]]
+
+# Handle missing values
+df = df.fillna(df.mean())
+
+# Downsample the majority class to handle class imbalance
+df_majority = df[df[class_col] == 0]
+df_minority = df[df[class_col] == 1]
+df_majority_downsampled = resample(df_majority, replace=False, n_samples=len(df_minority))
+df_balanced = pd.concat([df_majority_downsampled, df_minority])
+
+# Feature scaling
+X = df_balanced[feature_cols]
+y = df_balanced[class_col]
+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)
+
+# Balancing the dataset using SMOTE
+smote = SMOTE()
+X_train, y_train = smote.fit_resample(X_train, y_train)
+
+# Logistic Regression Model
+model = LogisticRegression(max_iter=1000)
+model.fit(X_train, y_train)
+
+# Predictions
+y_pred = model.predict(X_test)
+
+# Model evaluation
+print("Accuracy:", accuracy_score(y_test, y_pred))
+print("Classification Report:\n", classification_report(y_test, y_pred))
+
+# Initialize the retrieval pipeline with a lightweight model (if required)
+retrieval_pipeline = pipeline("feature-extraction", model="distilbert-base-uncased")
+
+def retrieve_explanation(prediction):
+    if prediction == 1:
+        explanation = "The transaction is classified as fraudulent based on the provided features."
+    else:
+        explanation = "The transaction is classified as non-fraudulent based on the provided features."
+    return explanation
+
+# Gradio prediction function with complete feature padding
+def fraud_detection_predictor(V1, V2, V3, Amount):
+    # Create a list of features with default zero values for missing ones
+    input_features = [0] * len(feature_cols)
+    
+    # Map the provided features to their indices (ensure they are in correct feature_cols)
+    v1_index = feature_cols.index('V1')  # Ensure these columns exist in feature_cols
+    v2_index = feature_cols.index('V2')
+    v3_index = feature_cols.index('V3')
+    amount_index = feature_cols.index('Amount')
+    
+    # Assign user inputs to the correct feature indices
+    input_features[v1_index] = V1
+    input_features[v2_index] = V2
+    input_features[v3_index] = V3
+    input_features[amount_index] = Amount
+
+    # Scale input data using the pre-fitted scaler
+    input_data = scaler.transform([input_features])
+    
+    # Make a prediction
+    prediction = model.predict(input_data)[0]
+    fraud_status = "Fraudulent" if prediction == 1 else "Non-Fraudulent"
+    
+    # Get explanation
+    explanation = retrieve_explanation(prediction)
+    return fraud_status, explanation
+
+# Define Gradio Interface
+interface = gr.Interface(
+    fn=fraud_detection_predictor,
+    inputs=[
+        gr.Number(label="V1"),
+        gr.Number(label="V2"),
+        gr.Number(label="V3"),
+        gr.Number(label="Amount")
+    ],
+    outputs=[
+        gr.Textbox(label="Fraud Status"),
+        gr.Textbox(label="Explanation")
+    ],
+    title="Simplified Credit Card Fraud Detection",
+    description="Enter a few transaction features (V1, V2, V3, Amount) to predict fraud status."
+)
+
+# Launch Gradio Interface
+interface.launch()