DiabetesPrediction / src /streamlit_app.py
EnYa32's picture
Update src/streamlit_app.py
7e92c15 verified
Raw History Blame Contribute Delete
7.2 kB
import numpy as np
import pandas as pd
import streamlit as st
import joblib
from pathlib import Path
# -----------------------------
# Page config
# -----------------------------
st.set_page_config(
page_title='Diabetes Risk Predictor (LightGBM)',
page_icon='🩺',
layout='centered'
)
st.title('🩺 Diabetes Risk Predictor')
st.write('Predicts the probability of being diagnosed with diabetes using a trained LightGBM model.')
# -----------------------------
# Paths (HF-friendly)
# -----------------------------
BASE_DIR = Path(__file__).resolve().parent
MODEL_PATH = BASE_DIR / 'lgbm_model.pkl'
FEATURE_COLS_PATH = BASE_DIR / 'feature_cols.pkl'
# -----------------------------
# Load artifacts
# -----------------------------
@st.cache_resource
def load_artifacts():
if not MODEL_PATH.exists():
raise FileNotFoundError(
f'Model not found: {MODEL_PATH.name}. Put it in the repo root (same folder as app.py).'
)
if not FEATURE_COLS_PATH.exists():
raise FileNotFoundError(
f'Feature list not found: {FEATURE_COLS_PATH.name}. Put it in the repo root (same folder as app.py).'
)
model = joblib.load(MODEL_PATH)
feature_cols = joblib.load(FEATURE_COLS_PATH)
return model, feature_cols
model, feature_cols = load_artifacts()
# -----------------------------
# Feature engineering (must match your notebook logic)
# -----------------------------
def add_features(df: pd.DataFrame) -> pd.DataFrame:
df = df.copy()
eps = 1e-6
# Body / Age
df['bmi_age_ratio'] = df['bmi'] / (df['age'] + eps)
df['is_obese'] = (df['bmi'] >= 30).astype(int)
# Activity
df['activity_per_age'] = df['physical_activity_minutes_per_week'] / (df['age'] + eps)
df['low_activity'] = (df['physical_activity_minutes_per_week'] < 150).astype(int)
# Sleep
df['sleep_deficit'] = 7 - df['sleep_hours_per_day']
df['poor_sleep'] = (df['sleep_hours_per_day'] < 6).astype(int)
# Cholesterol
if {'ldl_cholesterol', 'hdl_cholesterol'}.issubset(df.columns):
df['chol_ratio'] = df['ldl_cholesterol'] / (df['hdl_cholesterol'] + eps)
df['high_ldl'] = (df['ldl_cholesterol'] >= 130).astype(int)
df['low_hdl'] = (df['hdl_cholesterol'] < 40).astype(int)
return df
# -----------------------------
# UI Inputs
# -----------------------------
with st.form('input_form'):
st.subheader('Patient Information')
col1, col2 = st.columns(2)
with col1:
age = st.number_input('Age', min_value=1, max_value=120, value=45, step=1)
gender = st.selectbox('Gender', ['Female', 'Male', 'Other'])
ethnicity = st.selectbox('Ethnicity', ['White', 'Hispanic', 'Black', 'Asian', 'Other'])
education_level = st.selectbox('Education Level', ['Highschool', 'Graduate', 'Postgraduate', 'No formal'])
employment_status = st.selectbox('Employment Status', ['Employed', 'Retired', 'Unemployed', 'Student'])
with col2:
income_level = st.selectbox('Income Level', ['Low', 'Lower-Middle', 'Middle', 'Upper-Middle', 'High'])
smoking_status = st.selectbox('Smoking Status', ['Never', 'Former', 'Current'])
family_history_diabetes = st.selectbox('Family History of Diabetes', [0, 1], index=0)
hypertension_history = st.selectbox('Hypertension History', [0, 1], index=0)
cardiovascular_history = st.selectbox('Cardiovascular History', [0, 1], index=0)
st.subheader('Lifestyle & Health Metrics')
col3, col4 = st.columns(2)
with col3:
alcohol_consumption_per_week = st.number_input('Alcohol Consumption per Week', min_value=0, max_value=50, value=2, step=1)
physical_activity_minutes_per_week = st.number_input('Physical Activity (minutes/week)', min_value=0, max_value=2000, value=90, step=5)
diet_score = st.number_input('Diet Score', min_value=0.0, max_value=10.0, value=6.0, step=0.1)
sleep_hours_per_day = st.number_input('Sleep Hours per Day', min_value=0.0, max_value=24.0, value=7.0, step=0.1)
screen_time_hours_per_day = st.number_input('Screen Time Hours per Day', min_value=0.0, max_value=24.0, value=6.0, step=0.1)
with col4:
bmi = st.number_input('BMI', min_value=10.0, max_value=60.0, value=26.0, step=0.1)
waist_to_hip_ratio = st.number_input('Waist-to-Hip Ratio', min_value=0.50, max_value=1.50, value=0.86, step=0.01)
systolic_bp = st.number_input('Systolic BP', min_value=50, max_value=250, value=116, step=1)
diastolic_bp = st.number_input('Diastolic BP', min_value=30, max_value=160, value=75, step=1)
heart_rate = st.number_input('Heart Rate', min_value=20, max_value=220, value=70, step=1)
col5, col6 = st.columns(2)
with col5:
cholesterol_total = st.number_input('Total Cholesterol', min_value=50, max_value=400, value=187, step=1)
hdl_cholesterol = st.number_input('HDL Cholesterol', min_value=5, max_value=150, value=54, step=1)
with col6:
ldl_cholesterol = st.number_input('LDL Cholesterol', min_value=5, max_value=300, value=103, step=1)
triglycerides = st.number_input('Triglycerides', min_value=10, max_value=600, value=123, step=1)
submitted = st.form_submit_button('Predict Risk')
# -----------------------------
# Prediction
# -----------------------------
if submitted:
# Build one-row dataframe (must match your training columns BEFORE encoding)
row = {
'age': int(age),
'alcohol_consumption_per_week': int(alcohol_consumption_per_week),
'physical_activity_minutes_per_week': int(physical_activity_minutes_per_week),
'diet_score': float(diet_score),
'sleep_hours_per_day': float(sleep_hours_per_day),
'screen_time_hours_per_day': float(screen_time_hours_per_day),
'bmi': float(bmi),
'waist_to_hip_ratio': float(waist_to_hip_ratio),
'systolic_bp': int(systolic_bp),
'diastolic_bp': int(diastolic_bp),
'heart_rate': int(heart_rate),
'cholesterol_total': int(cholesterol_total),
'hdl_cholesterol': int(hdl_cholesterol),
'ldl_cholesterol': int(ldl_cholesterol),
'triglycerides': int(triglycerides),
'gender': gender,
'ethnicity': ethnicity,
'education_level': education_level,
'income_level': income_level,
'smoking_status': smoking_status,
'employment_status': employment_status,
'family_history_diabetes': int(family_history_diabetes),
'hypertension_history': int(hypertension_history),
'cardiovascular_history': int(cardiovascular_history),
}
X_user = pd.DataFrame([row])
# Feature engineering
X_user = add_features(X_user)
# One-hot encode
X_user_enc = pd.get_dummies(X_user, drop_first=True)
# Align columns to training feature set
X_user_enc = X_user_enc.reindex(columns=feature_cols, fill_value=0)
# Predict probability
proba = float(model.predict_proba(X_user_enc)[:, 1][0])
st.success(f'Predicted Diabetes Probability: {proba:.3f}')
st.progress(min(max(proba, 0.0), 1.0))
st.caption('Note: This app outputs probability (0–1). It is not a medical diagnosis.')