Spaces:
Sleeping
Sleeping
Upload app.py
Browse files
app.py
ADDED
|
@@ -0,0 +1,86 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
import streamlit as st
|
| 2 |
+
import pandas as pd
|
| 3 |
+
import numpy as np
|
| 4 |
+
import joblib
|
| 5 |
+
from xgboost import XGBRegressor
|
| 6 |
+
from sklearn.preprocessing import StandardScaler, OneHotEncoder
|
| 7 |
+
import shap
|
| 8 |
+
from streamlit_shap import st_shap
|
| 9 |
+
|
| 10 |
+
|
| 11 |
+
|
| 12 |
+
# Page configuration
|
| 13 |
+
st.set_page_config(
|
| 14 |
+
page_title="Medical Costs Concern Prediction",)
|
| 15 |
+
|
| 16 |
+
st.title('Predict Medical Costs Concern')
|
| 17 |
+
|
| 18 |
+
# Load model and preprocessing objects
|
| 19 |
+
@st.cache_resource
|
| 20 |
+
def load_model_objects():
|
| 21 |
+
model_xgb = joblib.load('model_best.joblib')
|
| 22 |
+
scaler = joblib.load('scaler.joblib')
|
| 23 |
+
return model_xgb, scaler
|
| 24 |
+
|
| 25 |
+
model_xgb, scaler = load_model_objects()
|
| 26 |
+
# Create SHAP explainer
|
| 27 |
+
explainer = shap.TreeExplainer(model_xgb)
|
| 28 |
+
|
| 29 |
+
# App description
|
| 30 |
+
with st.expander("What's this app?"):
|
| 31 |
+
st.markdown("""
|
| 32 |
+
This app predicts how worried a person is about medical costs, based on factors like age, education, income, and employment status.
|
| 33 |
+
We've trained an AI model to analyze these inputs and give a prediction.
|
| 34 |
+
""")
|
| 35 |
+
|
| 36 |
+
st.subheader('Describe yourself')
|
| 37 |
+
|
| 38 |
+
# User inputs
|
| 39 |
+
col1, col2 = st.columns(2)
|
| 40 |
+
|
| 41 |
+
with col1:
|
| 42 |
+
age = st.number_input('Age', min_value=18, max_value=100, value=30)
|
| 43 |
+
education = st.selectbox('Education Level', options=['Primary', 'Secondary', 'Tertiary'], index=1)
|
| 44 |
+
income_quartile = st.radio('Income Quartile', options=['Lowest', 'Second', 'Third', 'Highest'])
|
| 45 |
+
|
| 46 |
+
with col2:
|
| 47 |
+
employment_status = st.selectbox('Employment Status', options=['Unemployed', 'Employed'], index=1)
|
| 48 |
+
|
| 49 |
+
# Map user inputs to numerical and categorical features
|
| 50 |
+
education_mapping = {'Primary': 1, 'Secondary': 2, 'Tertiary': 3}
|
| 51 |
+
income_mapping = {'Lowest': 1, 'Second': 2, 'Third': 3, 'Highest': 4}
|
| 52 |
+
employment_mapping = {'Unemployed': 0, 'Employed': 1}
|
| 53 |
+
|
| 54 |
+
# Transform user input into a feature vector
|
| 55 |
+
education_num = education_mapping[education]
|
| 56 |
+
income_num = income_mapping[income_quartile]
|
| 57 |
+
employment_num = employment_mapping[employment_status]
|
| 58 |
+
|
| 59 |
+
# Prepare features for the model
|
| 60 |
+
num_features = pd.DataFrame({
|
| 61 |
+
'age': [age],
|
| 62 |
+
'educ': [education_num],
|
| 63 |
+
'inc_q': [income_num],
|
| 64 |
+
'emp_in': [employment_num]
|
| 65 |
+
})
|
| 66 |
+
num_scaled = pd.DataFrame(scaler.transform(num_features), columns=num_features.columns)
|
| 67 |
+
|
| 68 |
+
# Prediction button
|
| 69 |
+
if st.button('Predict Concern Level'):
|
| 70 |
+
# Make prediction
|
| 71 |
+
predicted_concern = model_xgb.predict(num_scaled)[0]
|
| 72 |
+
# Display prediction
|
| 73 |
+
st.metric(label="Predicted concern level", value=f'{round(predicted_concern)} (1: Not Worried, 3: Very Worried)')
|
| 74 |
+
# SHAP explanation
|
| 75 |
+
st.subheader('Concern Factors Explained')
|
| 76 |
+
shap_values = explainer.shap_values(num_scaled)
|
| 77 |
+
st_shap(shap.force_plot(explainer.expected_value, shap_values, num_scaled), height=400, width=600)
|
| 78 |
+
st.markdown("""
|
| 79 |
+
This plot shows how each feature contributes to the predicted concern level:
|
| 80 |
+
- Blue bars push the concern level lower
|
| 81 |
+
- Red bars push the concern level higher
|
| 82 |
+
- The length of each bar indicates the strength of the feature's impact
|
| 83 |
+
""")
|
| 84 |
+
|
| 85 |
+
# Footer
|
| 86 |
+
st.markdown("---")
|