app.py

import streamlit as st
import pandas as pd
import numpy as np
import torch
from predict import TabularTransformer, model_predict
from sklearn.preprocessing import MinMaxScaler
import matplotlib.pyplot as plt
import shap

# Set page config
st.set_page_config(
    page_title="Resistivity Prediction App",
    page_icon="🔮",
    layout="wide"
)

# Title and description
st.title("Resistivity Prediction App")
st.markdown("""
This app predicts resistivity based on input features. Enter the values for each feature
and click 'Predict' to get the prediction and explanation.
""")

@st.cache_resource
def load_model_and_scalers():
    # Load data for scaling
    df = pd.read_excel('data.xlsx')
    X = df.iloc[:, 0:8]
    y = df.iloc[:, 8]
    feature_names = X.columns.tolist()
    
    # Initialize scalers
    scaler_X = MinMaxScaler()
    scaler_y = MinMaxScaler()
    
    # Fit scalers
    scaler_X.fit(X)
    scaler_y.fit(y.values.reshape(-1, 1))
    
    # Load model
    model = TabularTransformer(input_dim=8, output_dim=1)
    model.load_state_dict(torch.load('model.pth'))
    model.eval()
    
    return model, scaler_X, scaler_y, feature_names, X

def explain_prediction(model, input_df, X_background, scaler_X, scaler_y, feature_names):
    # Create a prediction function for SHAP
    def predict_fn(X):
        X_tensor = torch.FloatTensor(scaler_X.transform(X))
        with torch.no_grad():
            scaled_pred = model(X_tensor).numpy()
        return scaler_y.inverse_transform(scaled_pred)
    
    # Use a subset of training data as background
    background_sample = X_background.sample(n=min(100, len(X_background)), random_state=42)
    explainer = shap.KernelExplainer(predict_fn, background_sample)
    
    # Calculate SHAP values for the input
    shap_values = explainer.shap_values(input_df)
    
    # Handle different SHAP value formats
    if isinstance(shap_values, list):
        shap_values = np.array(shap_values[0])
    
    # Ensure correct shape for waterfall plot
    if len(shap_values.shape) > 1:
        if shap_values.shape[0] == len(feature_names):
            shap_values = shap_values.T
        shap_values = shap_values.flatten()
    
    # Create waterfall plot
    plt.figure(figsize=(12, 8))
    shap.plots.waterfall(
        shap.Explanation(
            values=shap_values,
            base_values=explainer.expected_value if np.isscalar(explainer.expected_value) 
                       else explainer.expected_value[0],
            data=input_df.iloc[0].values,
            feature_names=feature_names
        ),
        show=False
    )
    plt.title('SHAP Value Contributions')
    plt.tight_layout()
    plt.savefig('shap_explanation.png', dpi=300, bbox_inches='tight')
    plt.close()
    
    return explainer.expected_value, shap_values

# Load model and scalers
try:
    model, scaler_X, scaler_y, feature_names, X = load_model_and_scalers()
    
    # Create input fields for features
    st.subheader("Input Features")
    
    # Create two columns for input fields
    col1, col2 = st.columns(2)
    
    # Dictionary to store input values
    input_values = {}
    
    # Create input fields split between two columns
    for i, feature in enumerate(feature_names):
        # Get min and max values for each feature
        min_val = float(X[feature].min())
        max_val = float(X[feature].max())
        
        # Add input field to alternating columns
        with col1 if i < len(feature_names)//2 else col2:
            input_values[feature] = st.number_input(
                f"{feature}",
                min_value=float(min_val),
                max_value=float(max_val),
                value=float(X[feature].mean()),
                help=f"Range: {min_val:.2f} to {max_val:.2f}"
            )
    
    # Add predict button
    if st.button("Predict"):
        # Create input DataFrame
        input_df = pd.DataFrame([input_values])
        
        # Make prediction
        prediction = model_predict(model, input_df, scaler_X, scaler_y)
        
        # Display prediction
        st.subheader("Prediction Result")
        st.markdown(f"### Predicted Resistivity: {prediction[0]:.2f}")
        
        # Calculate and display SHAP values
        st.subheader("Feature Importance Explanation")
        
        # Get SHAP values using the training data as background
        expected_value, shap_values = explain_prediction(
            model, input_df, X, scaler_X, scaler_y, feature_names
        )
        
        # Display the waterfall plot
        st.image('shap_explanation.png')

except Exception as e:
    st.error(f"""
    Error loading the model and data. Please make sure:
    1. The model file 'model.pth' exists
    2. The data file 'data.xlsx' exists
    3. All required packages are installed
    
    Error details: {str(e)}
    """)