DP1110/mlp-accessibility-model

1import pandas as pd2import joblib3from huggingface_hub import hf_hub_download4from sklearn.impute import SimpleImputer5import numpy as np6 7# Define the Hugging Face repository ID and filenames8REPO_ID = "DP1110/mlp-accessibility-model"9MODEL_FILENAME = 'mlp_regressor_model.joblib'10IMPUTER_FILENAME = 'simple_imputer.joblib'11 12# Define the feature columns, matching the training data order13FEATURE_COLUMNS = ['% ASF (Euclidean)', '% Built-Up Area', '% ASF (Network)', '% ASF from Bus Stops ', '% ASF from Bus Stops', '% ASF (Network) ']14 15# Download the model and imputer from Hugging Face Hub16try:17    model_path = hf_hub_download(repo_id=REPO_ID, filename=MODEL_FILENAME)18    imputer_path = hf_hub_download(repo_id=REPO_ID, filename=IMPUTER_FILENAME)19except Exception as e:20    print('Error downloading files from Hugging Face Hub:', e)21    model_path = None22    imputer_path = None23 24# Load the model and imputer25loaded_mlp_model = None26loaded_imputer = None27 28if model_path:29    loaded_mlp_model = joblib.load(model_path)30    print('MLP model loaded from', model_path)31 32if imputer_path:33    loaded_imputer = joblib.load(imputer_path)34    print('Imputer loaded from', imputer_path)35 36def predict_accessibility_score(new_data_df: pd.DataFrame) -> pd.Series:37    """38    Predicts the overall accessibility score for new, raw input data.39 40    Args:41        new_data_df (pd.DataFrame): A DataFrame containing new data with the same42                                    feature columns as the training data, before imputation.43 44    Returns:45        pd.Series: Predicted overall accessibility scores.46    """47    if loaded_mlp_model is None or loaded_imputer is None:48        raise RuntimeError('Model or imputer not loaded. Cannot make predictions.')49 50    # Ensure the order of columns matches the training data51    # Handle cases where new_data_df might have different columns or order52    missing_cols = set(FEATURE_COLUMNS) - set(new_data_df.columns)53    for c in missing_cols:54        new_data_df[c] = np.nan  # Or appropriate default value55 56    # Reorder columns to match the training features57    new_data_df = new_data_df[FEATURE_COLUMNS]58 59    # Apply the loaded imputer to handle missing values in new data60    new_data_imputed = loaded_imputer.transform(new_data_df)61    new_data_imputed_df = pd.DataFrame(new_data_imputed, columns=FEATURE_COLUMNS)62 63    # Make predictions using the loaded MLP model64    predictions = loaded_mlp_model.predict(new_data_imputed_df)65 66    return pd.Series(predictions, name='Predicted_Overall_Accessibility_Score')67 68if __name__ == '__main__':69    print("\n--- Demonstrating prediction with sample data ---")70 71    # Create a sample DataFrame for new raw data.72    # This should mimic the structure of the features used for training.73    sample_data_dict = {}74    for i, col_name in enumerate(FEATURE_COLUMNS):75        # Assign arbitrary values for demonstration76        sample_data_dict[col_name] = [0.5 + (i * 0.005) % 0.1]  # Varying slightly for demonstration77 78    new_sample_data = pd.DataFrame(sample_data_dict)79 80    # Make predictions using the defined function81    try:82        predictions = predict_accessibility_score(new_sample_data)83 84        # Display the new sample data and the predictions85        print("\n--- New Sample Data for Prediction ---")86        print(new_sample_data)87        print("\n--- Predicted Overall Accessibility Score ---")88        print(predictions)89    except Exception as e:90        print('Error during prediction:', e)91