reporting.py

"""
Python script meant to generate reports about model metrics.
"""
from sklearn import metrics
import matplotlib.pyplot as plt
import logging
import json
import os
import sys
from diagnostics import model_predictions

logging.basicConfig(stream=sys.stdout, level=logging.INFO)


with open('config.json', 'r') as f:
    """
    Load config.json and correct path variable.
    """
    config = json.load(f) 

model_path = os.path.join(config['output_model_path'])
test_data_path = os.path.join(config['test_data_path']) 


def score_model():
    """
    Function for reporting.
    """
    # Calculate a confusion matrix using the test data and the deployed model.
    # Write the confusion matrix to the workspace.
    y_pred, df_y = model_predictions(None)
    df_cm = metrics.confusion_matrix(df_y, y_pred)
    fig, ax = plt.subplots(figsize=(7.5, 7.5))
    ax.matshow(df_cm, alpha=0.3)
    for i in range(df_cm.shape[0]):
        for j in range(df_cm.shape[1]):
            ax.text(x=j, y=i, s=df_cm[i, j], va='center', ha='center', size='xx-large')

    plt.xlabel('Predictions', fontsize=18)
    plt.ylabel('Actuals', fontsize=18)
    plt.title('Confusion Matrix', fontsize=18)
    plt.savefig(os.path.join(model_path, "confusionmatrix.png"))


if __name__ == "__main__":
    logging.info("Running Reporting!")
    score_model()