app.py

import gradio as gr
import pandas as pd
import joblib
import numpy as np

# === Load trained MLP model for NHNLC prediction ===
model = joblib.load("MLP_LDL_Predict_by_HNHLC.s3a")  # Ensure model file is in the same directory

# === Prediction function ===
def predict_ldl(tc, tg, hdl):
    try:
        # Step 1: Derived inputs
        nhc = tc - hdl
        tg_hdl_ratio = tg / hdl if hdl != 0 else 0  # avoid divide-by-zero

        # Step 2: Prediction of NHNLC
        input_df = pd.DataFrame([[nhc, tg_hdl_ratio]], columns=["NHC", "TG/HDL"])
        nhnlc = model.predict(input_df)[0]

        # Step 3: LDL Calculation
        ldl = tc - hdl - nhnlc
        nhnlc = round(nhnlc, 0)
        ldl = round(ldl, 0)

        return f"The Non-HDL Non-LDL Cholesterol = {nhnlc} mg/dL\nThe final estimated LDL = {ldl} mg/dL"

    except Exception as e:
        return f"Error: {str(e)}"

# === Gradio UI ===
interface = gr.Interface(
    fn=predict_ldl,
    inputs=[
        gr.Number(label="Total Cholesterol (mg/dL)"),
        gr.Number(label="Triglycerides (mg/dL)"),
        gr.Number(label="HDL Cholesterol (mg/dL)")
    ],
    outputs=gr.Textbox(label="LDL Estimation Result"),
    title="MLP Neural Network LDL Estimator (via NHNLC)",
    description="This model estimates LDL using a 2-step approach: first it predicts Non-HDL Non-LDL Cholesterol (NHNLC) using NHC and TG/HDL ratio, then calculates LDL = TC - HDL - NHNLC."
)

# === Launch interface ===
interface.launch()