Create app.py

app.py

@@ -0,0 +1,106 @@
+import gradio as gr
+import requests
+import json
+import plotly
+
+def predict_fraud(selected_model, selected_interpretability_method, step, transaction_type, amount, oldbalanceOrg):
+    url = "https://sea-lion-app-5kxay.ondigitalocean.app"
+    data = {
+        'selected_model': selected_model,
+        'selected_interpretability_method': selected_interpretability_method,
+        'step': step,
+        'transaction_type': transaction_type,
+        'amount': amount,
+        'oldbalanceOrg': oldbalanceOrg
+    }
+
+    response = requests.post(url, json=data)
+    if response.status_code == 200:
+        result = response.json()
+
+        # Directly use the base64-encoded image string for the network graph
+        network_graph = result['network_graph']
+
+        # Ensure other data is handled correctly
+        prediction_text = result['prediction_text']
+        model_explanation = result['model_explanation']
+       
+
+
+        mod_plot_json = result['mod_plot']        
+        # Parse the JSON strings back into Plotly figures
+        mod_plot = plotly.graph_objs.Figure(json.loads(mod_plot_json))        
+        features_influence = result['features_influence']
+        
+        network_graph_json = result['network_graph']   #graph_objects 
+        # Parse the JSON strings back into Plotly figures        
+        network_graph = plotly.graph_objs.Figure(json.loads(network_graph_json))        
+        network_explainer = result['network_explainer']
+        top_main_effect = result['top_main_effect']
+        top_interaction = result['top_interaction']        
+        
+        # Parse the JSON strings back into Plotly figures        
+        radial_plot_json = result['radial_plot']
+        bar_chart_json = result['bar_chart']        
+        radial_plot = plotly.graph_objs.Figure(json.loads(radial_plot_json))
+        bar_chart = plotly.graph_objs.Figure(json.loads(bar_chart_json))
+                        
+        narrative = result.get('narrative', "")
+
+      # Return the results
+        return prediction_text, model_explanation, mod_plot, features_influence, network_graph, network_explainer, top_main_effect, top_interaction,radial_plot, bar_chart, narrative
+    else:
+        # Handle error scenario by returning placeholders for each expected output
+        return "Error: " + response.text, None, None, None, None,None, None, None, None, None, None
+   
+    
+# Define your Gradio interface here
+with gr.Blocks() as app:
+    gr.Markdown("<h2 style='text-align: center; font-weight: bold;'>FraudSenseXAI - Advanced Fraud Detection</h2>")
+    gr.Markdown("<p style='text-align: center;'>Predict and analyze fraudulent transactions.</p>", elem_id="description")
+    
+    
+    with gr.Row():
+        with gr.Column():
+            gr.Markdown("#### INPUT PARAMTERS")
+            model_selection = gr.Dropdown(choices=['Random Forest', 'Gradient Boost'], label="Model Selection")
+            interpretability_selection = gr.Dropdown(choices=['LIME', 'SHAP'], label="Interpretability Technique")
+            step = gr.Number(label="Step")
+            transaction_type = gr.Dropdown(choices=['Transfer', 'Payment', 'Cash Out', 'Cash In'], label="Transaction Type")
+            transaction_amount = gr.Number(label="Transaction Amount")
+            old_balance_org = gr.Number(label="Old Balance Org")
+            submit_btn = gr.Button("Analyze")
+    
+        # Define outputs
+            gr.Markdown("#### PREDICTION RESULT")
+            prediction_text = gr.Textbox(label="Prediction", lines=7)
+        
+        with gr.Column():
+            gr.Markdown("#### MODEL INTERPRETATIONS")
+            model_explanation = gr.Textbox(label="Model Explanation", lines=7)
+            
+            mod_plot = gr.Plot(label="Model Plot")
+            features_influence = gr.Textbox(label="Features Influence", lines=7)
+            
+            
+    with gr.Row():
+        with gr.Column():        
+            gr.Markdown("#### FEATURE INTERACTIONS: Note that this function only supports SHAP at the moment")     
+            network_graph = gr.Plot(label="Network Graph")
+            network_explainer = gr.Text(label="Network Graph Explanation")
+            top_main_effect = gr.Text(label="Top Main Effect", lines=7)
+            top_interaction = gr.Text(label="Top Interaction", lines=7)  
+        
+        with gr.Column():                  
+            gr.Markdown("#### COUNTERFACTUAL EXPLANATIONS")
+            radial_plot = gr.Plot(label="Radial Plot")
+            bar_chart = gr.Plot(label="Bar Chart")
+            narrative = gr.Textbox(label="Narrative")
+
+    submit_btn.click(
+        predict_fraud,
+        inputs=[model_selection, interpretability_selection, step, transaction_type, transaction_amount, old_balance_org],
+        outputs=[prediction_text, model_explanation, mod_plot, features_influence, network_graph, network_explainer, top_main_effect, top_interaction,radial_plot, bar_chart, narrative]
+    )
+
+app.launch(share=True)