app.py

import pickle
import pandas as pd
import sklearn
import gradio as gr
import joblib

def encode_df(df):
    
    #Gender
    sex_map = {"Male": 1, "Female": 0}
    df = df.replace({"Gender": sex_map})
    
    #Tipe Angina
    chestpain_map = {"Typical": 0, "Asymptomatic": 1, "Nonanginal": 2, "Nontypical" : 3}
    df = df.replace({"Tipe_angina": chestpain_map})
    
    #Gula_darah_puasa
    fastingBLP_map = {"<120": 0, ">120": 1, "120": 2}
    df = df.replace({"Gula_darah_puasa": fastingBLP_map})
    
    #Angina_aktivitas
    exang_map = {"No": 0, "Yes": 1}
    df = df.replace({"Angina_aktivitas": exang_map})
    
    #Hasil_Elektrokardiografi
    recg_map = {"Normal": 0, "Abnormal Wave": 1, "Left Ventricular Hypertrophy": 2}
    df = df.replace({"Hasil_Elektrokardiografi": recg_map})
    
    #st_slope_ECG
    slope_map = {"Upsloping": 1, "Flat": 2, "Downsloping": 3}
    df = df.replace({"st_slope_ECG": slope_map})
    
    df = df[
        [
            "Umur",
            "Gender",
            "Tipe_angina",
            "Tekanan_darah_istirahat",
            "Kolesterol",
            "Gula_darah_puasa",
            "Hasil_Elektrokardiografi",
            "Denyut_jantung_max",
            "Angina_aktivitas",
            "st_depression_ECG",
            "st_slope_ECG",
        ]
    ]
    return df

filename = 'ACS_model_random_forest.sav'

# load the model from disk
loaded_model = joblib.load(filename)

def predict(Umur, Gender, Tipe_angina, Tekanan_darah_istirahat, Kolesterol, 
            Gula_darah_puasa, Hasil_Elektrokardiografi, 
            Denyut_jantung_max, Angina_aktivitas, st_depression_ECG, st_slope_ECG):
    df = pd.DataFrame.from_dict(
        {
            "Umur": [Umur],
            "Gender": [Gender],
            "Tipe_angina": [Tipe_angina],
            "Tekanan_darah_istirahat" : [Tekanan_darah_istirahat],
            "Kolesterol" : [Kolesterol],
            "Gula_darah_puasa" : [Gula_darah_puasa],
            "Hasil_Elektrokardiografi": [Hasil_Elektrokardiografi],
            "Denyut_jantung_max": [Denyut_jantung_max],
            "Angina_aktivitas": [Angina_aktivitas],
            "st_depression_ECG": [st_depression_ECG],
            "st_slope_ECG": [st_slope_ECG],
        }
    )
    
    df = encode_df(df)
    pred = loaded_model.predict_proba(df)[0]
    return {"Possible Heart Disease": float(pred[1]), "Less chance of Heart Disease": float(pred[0])}

title = "Interactive Demonstration for Acute Coronary Syndrome Prediction System"
des = '''This model predicts the possibility of a heart disease using a hybrid sampling SMOTE-TOMEK model that achieved an high accuracy of 85%, and f1-score 94% with Random Forest Algorithm'''
article = "<p style='text-align: center'><a href='https://www.linkedin.com/in/m-afif-rizky-a-a96048182/'>Created by @Vrooh933 Production</a> | <a href='https://github.com/afifrizkyandika11551100310'>GitHub Profile</a>"

demo = gr.Interface(
    predict,
    [gr.Slider(0, 88, value=25, label='Umur'),
     gr.Radio(["Male", "Female"], label='Gender'),
     gr.Dropdown(["Typical", "Asymptomatic", "Nonanginal", "Nontypical"], label="Tipe_angina"),
     
     gr.Slider(0, 200, value=125, label= 'Tekanan_darah_istirahat'),
     gr.Slider(0, 603, value=50, label= 'Kolesterol'),
     gr.Radio(["<120", ">120", '120'], label='Gula_darah_puasa'),
     
     
     gr.Dropdown(["Normal", "Abnormal Wave", "Left Ventricular Hypertrophy"], label='Hasil_Elektrokardiografi'),
     gr.Number(value=100, label='Denyut_jantung_max'),
     gr.Dropdown(["No", "Yes"], label='Angina_aktivitas'),
     gr.Slider(-2.6, 6.2, value=3.1, label='st_depression_ECG'),
     gr.Radio(["Upsloping", "Flat", "Downsloping"], label='st_slope_ECG'),
    ],
    "label",
    examples=[
        [37,'Male','Typical', 89, 276, '>120', "Left Ventricular Hypertrophy",150,'No',2.3,'Downsloping',0,'Fixed'],
        [63,'Male','Typical', 50, 100, '<120', "Abnormal Wave",90,'No',1.0,'Downsloping',1,'Reversable'],
        [58,'Female','Asymptomatic', 70, 500, '>120', "Normal",111,'Yes',0.8,'Flat',0,'Normal']
    ],
    title=title,
    description=des,
    article=article,
    live=True,
)