File size: 4,721 Bytes
e0d8ab4 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 |
"""This module provides an explainer for the model."""
import shap
import pandas as pd
import numpy as np
class UhiExplainer:
"""
A class for SHAP-based model explanation.
Attributes:
- model: Trained model (e.g., RandomForestRegressor, XGBRegressor).
- explainer_type: SHAP explainer class (e.g., shap.TreeExplainer, shap.KernelExplainer).
- X: Data (Pandas DataFrame) used to compute SHAP values.
- feature_names: List of feature names.
- explainer: SHAP explainer instance.
- shap_values: Computed SHAP values.
Methods:
- apply_shap(): Computes SHAP values.
- summary_plot(): Generates a SHAP summary plot.
- bar_plot(): Generates a bar chart of feature importance.
- dependence_plot(): Generates a dependence plot for a feature.
- force_plot(): Generates a force plot for an individual prediction.
- init_js(): Initializes SHAP for Jupyter Notebook.
- reasoning(): Provides insights on why a record received a high or low UHI index.
"""
def __init__(self, model, explainer_type, X, feature_names, ref_data=None, shap_values=None):
"""
Initializes the Explainer with a trained model, explainer type, and dataset.
Parameters:
- model: Trained model (e.g., RandomForestRegressor, XGBRegressor).
- explainer_type: SHAP explainer class (e.g., shap.TreeExplainer, shap.KernelExplainer).
- X: Data (Pandas DataFrame) used to compute SHAP values.
- feature_names: List of feature names.
- ref_data (optional): The reference dataset (background dataset) is used by SHAP to estimate the expected output of the model
- shap_values (optional): Precomputed SHAP values
"""
self.model = model
self.explainer_type = explainer_type
self.X = np.array(X) if isinstance(X, pd.DataFrame) else X # Ensure NumPy format
if ref_data is not None:
ref_data = np.array(ref_data) if isinstance(ref_data, pd.DataFrame) else ref_data # Ensure NumPy format
self.feature_names = feature_names
self.explainer = explainer_type(model, ref_data) # Initialize explainer
# Compute SHAP values
if shap_values is not None:
self.shap_values = shap_values
else:
self.shap_values = self.explainer.shap_values(self.X, check_additivity=False) if self.explainer_type == shap.DeepExplainer else self.explainer.shap_values(self.X)
# Apply squeeze only if the array has three dimensions and the last dimension is 1
if self.shap_values.ndim == 3 and self.shap_values.shape[-1] == 1:
self.shap_values = np.squeeze(self.shap_values)
def reasoning(self, index=0, location=(None, None)):
"""
Provides insights on why the record received a high or low UHI index.
Parameters:
index (int): The index of the observation of interest.
location (tuple) (optional): The location of the record (long, lat).
Returns:
dict: The insights for the selected record.
"""
# Ensure expected_value is a single value (not tensor)
if self.explainer_type == shap.DeepExplainer:
expected_value = np.array(self.explainer.expected_value)
else:
expected_value = self.explainer.expected_value
# Extract single value if expected_value is an array
if isinstance(expected_value, np.ndarray):
expected_value = expected_value[0]
# Validate record index
if index >= len(self.shap_values) or index < 0:
return {"error": "Invalid record index"}
# Extract SHAP values for the specified record
record_shap_values = self.shap_values[index]
# Compute SHAP-based final prediction
shap_final_prediction = expected_value + sum(record_shap_values)
# Structure feature contributions
feature_contributions = [
{
"feature": feature,
"shap_value": value,
"impact": "increase" if value > 0 else "decrease"
}
for feature, value in zip(self.feature_names, record_shap_values)
]
# Create JSON structure
shap_json = {
"record_index": index,
"longitude": location[0],
"latitude": location[1],
"base_value": expected_value,
"shap_final_prediction": shap_final_prediction, # SHAP-based predicted value
"uhi_status": "Urban Heat Island" if shap_final_prediction > 1 else "Cooler Region",
"feature_contributions": feature_contributions,
}
return shap_json
|