app.py

import gradio as gr
import pandas as pd
import os
import joblib

models_folder = "models"

# Load pre-trained models
model_files = {
    "Random Forest": "RandomForest.pkl",
    "Decision Tree (CART)": "DecisionTree.pkl",
    "Support Vector Machine": "SVM.pkl",
    "K-Means (Clustering)": "KMeans.pkl",
    "Logistic Regression": "LogisticRegression.pkl",
    "XGBoost": "XGBoost.pkl",
    "Artificial Neural Network (ANN)": "ANN.pkl"
}

models = {
    name: joblib.load(os.path.join(models_folder, file))
    for name, file in model_files.items()
}

# BMI categorization function
def categorize_bmi(bmi):
    if bmi < 18:
        return 0
    elif 18 <= bmi <= 24:
        return 1
    elif 25 <= bmi <= 29:
        return 2
    elif 30 <= bmi <= 39:
        return 3
    else:
        return 4

# Prediction function
def predict_diabetes(
    model_name,
    HighBP,
    HighChol,
    CholCheck,
    Smoker,
    Stroke,
    HeartDiseaseorAttack,
    PhysActivity,
    Fruits,
    Veggies,
    HvyAlcoholConsump,
    AnyHealthcare,
    NoDocbcCost,
    GenHlth,
    MentHlth,
    PhysHlth,
    DiffWalk,
    Sex,
    Age,
    Education,
    Income,
    BMI
):
    model = models[model_name]

    binary_map = {"Yes": 1, "No": 0}
    sex_map = {"Male": 1, "Female": 0}

    # Create a dataframe from user inputs
    input_data = pd.DataFrame([{
        "HighBP": binary_map[HighBP],
        "HighChol": binary_map[HighChol],
        "CholCheck": binary_map[CholCheck],
        "Smoker": binary_map[Smoker],
        "Stroke": binary_map[Stroke],
        "HeartDiseaseorAttack": binary_map[HeartDiseaseorAttack],
        "PhysActivity": binary_map[PhysActivity],
        "Fruits": binary_map[Fruits],
        "Veggies": binary_map[Veggies],
        "HvyAlcoholConsump": binary_map[HvyAlcoholConsump],
        "AnyHealthcare": binary_map[AnyHealthcare],
        "NoDocbcCost": binary_map[NoDocbcCost],
        "GenHlth": GenHlth,
        "MentHlth": MentHlth,
        "PhysHlth": PhysHlth,
        "DiffWalk": binary_map[DiffWalk],
        "Sex": sex_map[Sex],
        "Age": Age,
        "Education": Education,
        "Income": Income,
        "BMI": categorize_bmi(BMI)
    }])

    # Convert specific columns to categorical types
    categorical_columns = ['GenHlth', 'Age', 'BMI', 'Education', 'Income']

    for col in categorical_columns:
        input_data[col] = input_data[col].astype('category')

    # Ensure all necessary columns exist, filling with 0 if missing
    expected_columns = pd.read_csv('dataset/diabetes_binary_5050split_health_indicators_BRFSS2015.csv').drop(columns=['Diabetes_binary']).columns
    input_data = input_data.reindex(columns=expected_columns, fill_value=0)

    # Make prediction
    if model_name == "K-Means (Clustering)":
        prediction = model.predict(input_data)
        prediction = ['At Risk' if p == 1 else 'Not at Risk' for p in prediction]
    else:
        prediction = model.predict(input_data)
        print(prediction)
        prediction = ['At Risk' if p == 1 else 'Not at Risk' for p in prediction]

    return prediction[0]

# Gradio UI with manual input fields
model_options = list(models.keys())

inputs = [
    gr.Dropdown(choices=model_options, label="Select Model", value="Random Forest"),
    gr.Radio(choices=["No", "Yes"], label="HighBP", value="No", info="Do you have High Blood Pressure?"),
    gr.Radio(choices=["No", "Yes"], label="HighChol", value="No", info="Do you have High Cholestrerol?"),
    gr.Radio(choices=["No", "Yes"], label="CholCheck", value="Yes", info="Have you had your cholesterol checked in past 5 years?"),
    gr.Radio(choices=["No", "Yes"], label="Smoker", value="No", info="Have you have smoked 100 cigarettes in entire life?"),
    gr.Radio(choices=["No", "Yes"], label="Stroke", value="No", info="Ever told you had a stroke?"),
    gr.Radio(choices=["No", "Yes"], label="HeartDiseaseorAttack", value="No", info="Have you ever had a heart disease or heart attack?"),
    gr.Radio(choices=["No", "Yes"], label="PhysActivity", value="Yes", info="Physical activity in past 30 days?"),
    gr.Radio(choices=["No", "Yes"], label="Fruits", value="Yes", info="Consumed fruit 1 or more times per day?"),
    gr.Radio(choices=["No", "Yes"], label="Veggies", value="Yes", info="Consumed veggies 1 or more times per day?"),
    gr.Radio(choices=["No", "Yes"], label="HvyAlcoholConsump", value="No", info="Adult men >= 14 drinks per week, Adult women >= 7 drinks per week"),
    gr.Radio(choices=["No", "Yes"], label="AnyHealthcare", value="Yes", info="Have any kind of healthcare coverage?"),
    gr.Radio(choices=["No", "Yes"], label="NoDocbcCost", value="No", info="Any time in past year needed a doctor but could not go because of cost?"),
    gr.Slider(1, 5, step=1, label="GenHlth", value=3, info="Rate general health 1-5 (1=excellent and 5=poor)"),
    gr.Slider(0, 30, step=1, label="MentHlth", value=0, info="Days of poor mental Health in past 30 days"),
    gr.Slider(0, 30, step=1, label="PhysHlth", value=0, info="Days of physcial illness or injury in past 30 days"),
    gr.Radio(choices=["No", "Yes"], label="DiffWalk", value="No", info="Difficulty walking"),
    gr.Radio(choices=["Male", "Female"], label="Sex", value="Female"),
    gr.Slider(1, 13, step=1, label="Age", value=7, info="Scale 1-13, 1=18-24 increments every 5 years 13=80+"),
    gr.Slider(1, 6, step=1, label="Education", value=4, info="Highest grate or year of school completed\n1=Never Attended\n2=Elementary\n3=Some Highschool\n4=Graduated Highschool\n5=Some College\n6=Graduated College"),
    gr.Slider(1, 8, step=1, label="Income", value=4, info="Scale 1-8 increments every $5k, 1=less than $10k\n5=$20-$35k\n8=$75k+"),
    gr.Slider(10, 50, step=1, label="BMI", value=25),
]


gr.Interface(
    fn=predict_diabetes,
    inputs=inputs,
    outputs="text",
    title="Diabetes Risk Classification",
    description="Enter health indicators to classify diabetes risk using different models."
).launch()