ACA050's picture
Update app.py
32b3395 verified
import pandas as pd
import shap
import matplotlib.pyplot as plt
import numpy as np
import joblib
import gradio as gr
# Load the model
try:
loaded_model = joblib.load('machine_failure_prediction_model.joblib')
print("Model loaded successfully for Gradio.")
except FileNotFoundError:
print("Error: 'machine_failure_prediction_model.joblib' not found. Ensure the model file is in the correct directory.")
loaded_model = None
def predict_failure_with_shap(Type, air_temp, process_temp, rotational_speed, torque, tool_wear):
if loaded_model is None:
return "Error: Model not loaded.", "", None
input_data = pd.DataFrame({
'Type': [Type],
'Air temperature [K]': [air_temp],
'Process temperature [K]': [process_temp],
'Rotational speed [rpm]': [rotational_speed],
'Torque [Nm]': [torque],
'Tool wear [min]': [tool_wear]
})
predicted_failure = loaded_model.predict(input_data)[0]
predicted_proba = loaded_model.predict_proba(input_data)[0]
prediction_label = f"Predicted Failure: {'Failure' if predicted_failure == 1 else 'No Failure'}"
probabilities_label = f"Probabilities (No Failure, Failure): {predicted_proba[0]:.4f}, {predicted_proba[1]:.4f}"
preprocessor = loaded_model.named_steps['preprocessor']
classifier = loaded_model.named_steps['classifier']
X_transformed = preprocessor.transform(input_data)
explainer = shap.TreeExplainer(classifier)
shap_values = explainer.shap_values(X_transformed)
feature_names = preprocessor.get_feature_names_out()
if isinstance(shap_values, list):
shap_val = shap_values[1][0]
base_val = explainer.expected_value[1]
else:
shap_val = shap_values[0, :] if shap_values.ndim == 2 else shap_values[0, :, 1]
base_val = explainer.expected_value if not isinstance(explainer.expected_value, (list, tuple, np.ndarray)) else explainer.expected_value[1]
fig = plt.figure()
shap.waterfall_plot(
shap.Explanation(
values=shap_val,
base_values=base_val,
data=X_transformed[0],
feature_names=feature_names
), show=False
)
plt.title("SHAP Waterfall Plot for Failure Prediction")
plt.tight_layout()
return prediction_label, probabilities_label, fig
# Define Gradio interface OUTSIDE the function
iface = gr.Interface(
fn=predict_failure_with_shap,
inputs=[
gr.Dropdown(choices=['L', 'M', 'H'], label='Machine Type'),
gr.Number(label='Air temperature [K]', step=0.1),
gr.Number(label='Process temperature [K]', step=0.1),
gr.Number(label='Rotational speed [rpm]', step=1),
gr.Number(label='Torque [Nm]', step=0.1),
gr.Number(label='Tool wear [min]', step=1)
],
outputs=[
gr.Label(label='Predicted Failure (0=No Failure, 1=Failure)'),
gr.Label(label='Prediction Probabilities'),
gr.Plot(label='SHAP Waterfall Plot')
],
title="Machine Failure Prediction with SHAP Analysis",
description="Enter the machine parameters to predict failure and see the SHAP analysis.",
flagging_mode='never'
)
if __name__ == "__main__":
iface.launch(server_name="0.0.0.0", server_port=7860)