app.py

import pandas as pd
import numpy as np
import gradio as gr


import os

from sklearn.ensemble import GradientBoostingRegressor

import joblib

from sklearn.impute import SimpleImputer

from sklearn.pipeline import Pipeline
from sklearn.compose import ColumnTransformer
from sklearn.preprocessing import (
    MinMaxScaler,
    OneHotEncoder,
    OrdinalEncoder,
    StandardScaler,
    PolynomialFeatures,
    FunctionTransformer,
)

#Load preprocessing and model pipelines
preprocessor_file = os.path.join('model', 'preprocessor.pkl') #Parámetros
preprocessor = joblib.load(preprocessor_file)

model_file = os.path.join('model', 'model.pkl') #Parámetros
model = joblib.load(model_file)

explainer_file = os.path.join('model', 'shapExplainer.pkl') #Parámetros
explainer = joblib.load(preprocessor_file)

# Function to make prediction and generate SHAP force plot
def predict_and_plot(X_test):
    X_test = preprocessor['preprocessor'].transform(X_test)
    prediction = model.predict(X_test)[0]
    
    # Generate SHAP force plot
    shap_values_single = explainer(input_array)
    force_plot_html = shap.force_plot(explainer.expected_value, shap_values_single.values, input_array, matplotlib=False)
    
    # Save the SHAP force plot to an HTML file
    shap_html_path = "force_plot.html"
    shap.save_html(shap_html_path, force_plot_html)
    
    return prediction, shap_html_path

# Define the Gradio interface
# Define the Gradio interface
input_df = gr.Dataframe(headers=preprocessor['cols'], row_count=1, col_count=len(preprocessor['cols']))
iface = gr.Interface(
    fn=predict_and_plot,
    inputs=input_df,
    outputs=[gr.Textbox(label="Predicted Value"), gr.HTML(label="SHAP Force Plot")],
    live=True
)

# Launch the Gradio app
iface.launch(share=True)