import gradio as gr import pandas as pd import numpy as np import joblib from sklearn.preprocessing import LabelEncoder, StandardScaler # Load model and feature columns model, feature_columns = joblib.load("delay_predictor_rf.pkl") encoders = joblib.load("encoders.pkl") scaler = joblib.load("scaler.pkl") # Define prediction function def predict_delay(role_sender, role_receiver, unit_sender, unit_receiver, message_type, classification_level, urgency_level, message_length_kb, expected_response_time): input_dict = { "role_sender": role_sender, "role_receiver": role_receiver, "unit_sender": unit_sender, "unit_receiver": unit_receiver, "message_type": message_type, "classification_level": classification_level, "urgency_level": urgency_level, "message_length_kb": float(message_length_kb), "expected_response_time": float(expected_response_time) } input_df = pd.DataFrame([input_dict]) # Encode categorical for col in encoders: input_df[col] = encoders[col].transform(input_df[col]) # Scale numerical input_df[["message_length_kb", "expected_response_time"]] = scaler.transform( input_df[["message_length_kb", "expected_response_time"]] ) # Reorder columns and enforce data types to match training input_df = input_df[feature_columns].copy() for col in input_df.columns: input_df[col] = input_df[col].astype(np.float64) # Predict and inverse log transform log_delay = model.predict(input_df)[0] delay_seconds = np.expm1(log_delay) return f"Estimated Response Delay: {delay_seconds:.2f} seconds (~{delay_seconds/60:.1f} minutes)" # Define Gradio interface iface = gr.Interface( fn=predict_delay, inputs=[ gr.Dropdown(choices=encoders["role_sender"].classes_.tolist(), label="Role Sender"), gr.Dropdown(choices=encoders["role_receiver"].classes_.tolist(), label="Role Receiver"), gr.Dropdown(choices=encoders["unit_sender"].classes_.tolist(), label="Unit Sender"), gr.Dropdown(choices=encoders["unit_receiver"].classes_.tolist(), label="Unit Receiver"), gr.Dropdown(choices=encoders["message_type"].classes_.tolist(), label="Message Type"), gr.Dropdown(choices=encoders["classification_level"].classes_.tolist(), label="Classification Level"), gr.Dropdown(choices=encoders["urgency_level"].classes_.tolist(), label="Urgency Level"), gr.Number(label="Message Length (KB)"), gr.Number(label="Expected Response Time (seconds)") ], outputs=gr.Text(label="Predicted Delay"), title="Military Communication Delay Predictor", description="Estimate response delay based on message metadata." ) # Launch app if __name__ == "__main__": iface.launch()