app.py

import gradio as gr
import joblib
import numpy as np
import pandas as pd
from models.svr import NuSVRInsuranceModel
from models.randomforest import RandomForestInsuranceModel
from models.XGBoost import XGBoostInsuranceModel
from models.knn import KNNInsuranceModel
from models.linreg import LinearRegressionInsuranceModel
from models.polyreg import PolynomialRegressionInsuranceModel
import os
# -------------------------------------------------------
# Placeholder code to load pre-trained models 
# (Replace 'model_NuSVR.joblib', etc. with your actual file paths)
# -------------------------------------------------------
model_nusvr = joblib.load(os.path.join('.', 'NuSVRInsuranceModel.joblib'))
model_xgb = joblib.load(os.path.join('.', 'XGBoostInsuranceModel.joblib'))
model_knn   = joblib.load(os.path.join('.', 'KNNInsuranceModel.joblib'))
model_rf    = joblib.load(os.path.join('.', 'RandomForestInsuranceModel.joblib'))
model_lr = joblib.load(os.path.join('.', 'LinearRegressionInsuranceModel.joblib'))
model_pr = joblib.load(os.path.join('.', 'PolynomialRegressionInsuranceModel.joblib'))

# # Dictionary to map model choice to the actual loaded model
models_dict = {
     "NuSVR": model_nusvr,
     "Gradient Boost Regressor": model_xgb,
     "KNN": model_knn,
     "Random Forest": model_rf,
     "Linear Regression": model_lr,
     "Polynomial Regression": model_pr
 }

# -------------------------------------------------------
# Simple label encodings or numeric mappings:
#   - gender: {"male":0, "female":1}
#   - diabetic: {"No":0, "Yes":1}
#   - smoker: {"No":0, "Yes":1}
#   - region: {"southwest":0, "southeast":1, "northwest":2, "northeast":3}
# Adjust this to however you pre-processed the data for training!
# -------------------------------------------------------
region_mapping = {"southwest": 0, "southeast": 1, "northwest": 2, "northeast": 3}
gender_mapping = {"male": 0, "female": 1}
yes_no_mapping = {"No": 0, "Yes": 1}

def predict_insurance_claim(
    model_choice, 
    age, 
    gender, 
    bmi, 
    blood_pressure, 
    diabetic, 
    children, 
    smoker, 
    region
):
    # Convert categorical values using the same logic used during training
    try:
        gender_val   = gender_mapping[gender.lower()]
        diabetic_val = yes_no_mapping[diabetic]
        smoker_val   = yes_no_mapping[smoker]
        region_val   = region_mapping[region.lower()]
    except KeyError:
        return "Invalid input for categorical field."

    # Construct a single-row DataFrame with the correct column names
    user_data = {
        'age':           [age],
        'gender':        [gender],
        'bmi':           [bmi],
        'bloodpressure': [blood_pressure],
        'diabetic':      [diabetic],
        'children':      [children],
        'smoker':        [smoker],
        'region':        [region]
    }

    user_df = pd.DataFrame(user_data)

    chosen_model = models_dict[model_choice]

    d = chosen_model.preprocessing(user_df)

    y_pred = chosen_model.predict(d)

    return float(y_pred[0])

# -------------------------------------------------------
# Gradio Interface
# -------------------------------------------------------
def build_interface():
    # Dropdown to choose model
    with gr.Row():
        model_dropdown = gr.Dropdown(
            choices=["NuSVR", "Gradient Boost Regressor", "KNN", "Random Forest", 'Linear Regression', 'Polynomial Regression'], 
            value="NuSVR", 
            label="Select Model"
        )
    
    with gr.Row():
    # Numeric inputs
        age_input = gr.Number(value=39.0, label="Age")
        bmi_input = gr.Number(value=23.2, label="BMI")
        bp_input  = gr.Number(value=91.0, label="Blood Pressure")
        children_input = gr.Number(value=0, label="Children")
    
        # Dropdowns for categorical data
        gender_input = gr.Dropdown(choices=["male", "female"], value="male", label="Gender")
        diabetic_input = gr.Dropdown(choices=["No", "Yes"], value="Yes", label="Diabetic")
        smoker_input = gr.Dropdown(choices=["No", "Yes"], value="No", label="Smoker")
        region_input = gr.Dropdown(choices=["southwest", "southeast", "northwest", "northeast"], 
                               value="southeast", 
                               label="Region")
    
    # Output
    output_label = gr.Textbox(label="Predicted Claim")
    
    # Interface
    demo = gr.Interface(
        fn=predict_insurance_claim,
        inputs=[
            model_dropdown,
            age_input,
            gender_input,
            bmi_input,
            bp_input,
            diabetic_input,
            children_input,
            smoker_input,
            region_input
        ],
        outputs=output_label,
        title="Insurance Claim Prediction"
    )
    
    return demo

if __name__ == "__main__":
    interface = build_interface()
    interface.launch()