File size: 8,274 Bytes
72d9d59
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
# polynomial_regression_model.py

import pandas as pd
import numpy as np
import joblib

from sklearn.preprocessing import OneHotEncoder, StandardScaler, PolynomialFeatures
from sklearn.compose import ColumnTransformer
from sklearn.pipeline import Pipeline
from sklearn.model_selection import train_test_split, GridSearchCV, cross_val_score
from sklearn.linear_model import LinearRegression
from sklearn.metrics import mean_absolute_error, r2_score

class PolynomialRegressionInsuranceModel:
    """
    A Polynomial Regression-based insurance claim prediction model with:
      1. Data loading & cleaning
      2. Preprocessing (categorical encoding, numerical scaling, polynomial features)
      3. Model training and evaluation
      4. Consistent API: preprocessing, predict, postprocessing
    """

    def __init__(self, csv_path):
        """
        Initializes the model by loading data, preprocessing, training, and evaluating.
        
        Parameters:
            csv_path (str): Path to the cleaned insurance data CSV file.
        """
        # -----------------------------------------------------
        # 1. Load and clean the data
        # -----------------------------------------------------
        df = pd.read_csv(csv_path)
        # Drop irrelevant columns and handle missing values
        df = df.drop(columns=['index', 'PatientID'], errors='ignore').dropna()

        # -----------------------------------------------------
        # 2. Handle outliers in the target variable 'claim'
        # -----------------------------------------------------
        target_column = 'claim'
        mean_y = df[target_column].mean()
        std_y = df[target_column].std()
        threshold_low = mean_y - 3.5 * std_y
        threshold_high = mean_y + 3.5 * std_y
        df = df[(df[target_column] >= threshold_low) & (df[target_column] <= threshold_high)]

        # -----------------------------------------------------
        # 3. Define features and target
        # -----------------------------------------------------
        self.features = df.drop(columns=[target_column])
        self.target = df[target_column].values  # or df['claim'].to_numpy()

        # -----------------------------------------------------
        # 4. Define preprocessing pipelines
        # -----------------------------------------------------
        categorical_columns = ['gender', 'smoker', 'region', 'diabetic']
        numerical_columns = ['bmi', 'bloodpressure', 'children', 'age']

        # Pipeline for categorical features
        categorical_pipeline = Pipeline([
            ('onehot', OneHotEncoder(handle_unknown='ignore'))
        ])

        # Pipeline for numerical features
        numerical_pipeline = Pipeline([
            ('scaler', StandardScaler())
        ])

        # Combine pipelines using ColumnTransformer
        self.preprocessor = ColumnTransformer([
            ('categorical', categorical_pipeline, categorical_columns),
            ('numerical', numerical_pipeline, numerical_columns)
        ])

        # Pipeline for polynomial features
        self.poly = PolynomialFeatures(degree=2, interaction_only=False, include_bias=False)

        # -----------------------------------------------------
        # 5. Combine preprocessing and polynomial features
        # -----------------------------------------------------
        self.full_preprocessor = Pipeline([
            ('preprocessor', self.preprocessor),
            ('poly', self.poly)
        ])

        # Transform the features
        X_preprocessed = self.full_preprocessor.fit_transform(self.features)

        # -----------------------------------------------------
        # 6. Train-test split
        # -----------------------------------------------------
        X_train, X_test, y_train, y_test = train_test_split(
            X_preprocessed,
            self.target,
            test_size=0.2,
            random_state=42
        )

        # -----------------------------------------------------
        # 7. Initialize and train the Linear Regression model
        # -----------------------------------------------------
        self.model = LinearRegression()

        # Perform 5-fold cross-validation on training data
        cv_scores = cross_val_score(self.model, X_train, y_train, cv=5, scoring='r2')
        print(f"[Polynomial Regression] Cross-Validation R2 Scores: {cv_scores}")
        print(f"[Polynomial Regression] Average CV R2 Score: {cv_scores.mean():.3f}")

        # Train the model on the full training data
        self.model.fit(X_train, y_train)

        # -----------------------------------------------------
        # 8. Evaluate the model on the test set
        # -----------------------------------------------------
        y_pred = self.model.predict(X_test)
        mae = mean_absolute_error(y_test, y_pred)
        r2 = r2_score(y_test, y_pred)
        self.__scores = {
            'MAE': mae,
            'R2': r2,
            'Cross-Validation R2 Scores': cv_scores,
            'Average CV R2': cv_scores.mean()
        }

        print(f"[Polynomial Regression] Test MAE: {mae:.3f}")
        print(f"[Polynomial Regression] Test R^2: {r2:.3f}")

    def preprocessing(self, raw_df):
        """
        Preprocesses new raw data by applying the same transformations as training.
        
        Parameters:
            raw_df (pd.DataFrame): New data with the same feature columns as training (excluding 'claim').
        
        Returns:
            np.ndarray: Transformed feature matrix ready for prediction.
        """
        return self.full_preprocessor.transform(raw_df)

    def predict(self, preprocessed_data):
        """
        Makes predictions on preprocessed data.
        
        Parameters:
            preprocessed_data (np.ndarray): Transformed feature matrix.
        
        Returns:
            np.ndarray: Predicted claim amounts.
        """
        preds = self.model.predict(preprocessed_data)
        return self.postprocessing(preds)

    def postprocessing(self, preds):
        """
        Postprocesses predictions. Currently a pass-through, but can be extended.
        
        Parameters:
            preds (np.ndarray): Raw predictions from the model.
        
        Returns:
            np.ndarray: Final predictions.
        """
        return preds

    def get_scores(self):
        """
        Retrieves the evaluation metrics.
        
        Returns:
            dict: Dictionary containing MAE, R2, and cross-validation scores.
        """
        return self.__scores

    def get_coefficients(self):
        """
        Retrieves the model's coefficients.
        
        Returns:
            pd.DataFrame: DataFrame of feature coefficients.
        """
        # Extract feature names after preprocessing and polynomial transformation
        categorical_features = self.preprocessor.named_transformers_['categorical'].named_steps['onehot'].get_feature_names_out(['gender', 'smoker', 'region', 'diabetic'])
        numerical_features = self.preprocessor.named_transformers_['numerical'].named_steps['scaler'].get_feature_names_out(['bmi', 'bloodpressure', 'children', 'age'])
        all_features = np.concatenate([categorical_features, numerical_features])

        # Get feature names after polynomial transformation
        poly_feature_names = self.poly.get_feature_names_out(all_features)

        # Create DataFrame of coefficients
        coefficients = pd.DataFrame({
            'Feature': poly_feature_names,
            'Coefficient': self.model.coef_
        }).sort_values(by='Coefficient', ascending=False)

        return coefficients

if __name__ == "__main__":
    # -----------------------------------------------------
    # 9. Instantiate and train the model
    # -----------------------------------------------------
    model = PolynomialRegressionInsuranceModel("cleaned_insurance_data.csv")

    # -----------------------------------------------------
    # 10. Export the entire model class instance
    # -----------------------------------------------------
    joblib.dump(model, "PolynomialRegressionInsuranceModel.joblib")
    print("Exported PolynomialRegressionInsuranceModel to PolynomialRegressionInsuranceModel.joblib")