import gradio as gr
import joblib
import numpy as np

# Load the trained model
model_path = 'RDF_Model_Classify_Severity_CKD.s3a'
model = joblib.load(model_path)

# Define the prediction function
def predict_ckd_severity(age, c3, c4):
    c3_c4_ratio = round(c3 / c4, 2)
    features = np.array([[age, c3, c4, c3_c4_ratio]])
    prediction = model.predict(features)[0]
    probability = model.predict_proba(features)[0][prediction] * 100
    
    if probability < 50:
        adjusted_probability = (probability / 50) * 100
    else:
        adjusted_probability = ((probability - 50) / 50) * 100
    
    
    
    if prediction == 1:
        result = f"The CKD status is most likely to progress into Severe (1). The probability is {adjusted_probability:.2f}%."
    else:
        result = f"The CKD status is most likely to progress into Non-Severe (0). The probability is {adjusted_probability:.2f}%."
    
    return c3_c4_ratio, result

# Create the Gradio interface
inputs = [
    gr.Number(label="Age (in years)"),
    gr.Number(label="C3"),
    gr.Number(label="C4")
]

outputs = [
    gr.Textbox(label="C3/C4 Ratio"),
    gr.Textbox(label="Prediction")
]

gr.Interface(fn=predict_ckd_severity, inputs=inputs, outputs=outputs, title="CKD Severity Prediction").launch()