balance_preditctor.py 文件源码

def build_regression_report(report_name, relevant_col_names, training_data, training_answers):
    '''
        Given a report_name, a list of columns to regress on, and the required training_data
        create a regression model using StatsModel. Plot the residuals and a QQ plot and write
        the model.summary() to the report.

        report_name: The name of the pdf
        relevant_col_names: a list with the columns you care about in training_data
        training_data: the training set
        training_answers: y, assumed to be parallel to training_data
    '''
    report = PdfPages(OUTPUT_DIR + report_name + '.pdf')
    reduced_dataset = training_data.filter(relevant_col_names)

    model, data = build_model(training_answers, reduced_dataset)
    summary_text = model.summary()
    with open(OUTPUT_DIR + report_name + ".txt", "w") as text_file:
        text_file.write(str(summary_text))

    resid_fig = plot_resid(model, data)
    report.savefig(resid_fig)
    report.close()