Loan default notebook¶

!pip install datarobot umap-learn nbformat hdbscan

import datarobot as dr
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import matplotlib.ticker as mtick

light_blue = "#598fd6"
grey_blue = "#5f728b"
orange = "#dd6b3d"

# Create the project and upload data
project = dr.Project.create(sourcedata = pathfinder_df,
                            project_name = 'Predict loan defaults')

# Set the project target to the appropriate feature. Use the LogLoss metric to measure performance and specify Autopilot to run on a subset of models
project.set_target(target='is_bad',
                  mode=dr.AUTOPILOT_MODE.QUICK, 
                  worker_count='-1')

# Uncomment and replace the project ID if the project already exists
# project = dr.Project.get("612cb904ce5d5617d67af394")

# Get the project metric (i.e LogLoss, RMSE, etc...)
metric = project.metric

# Get project URL
project_url = project.get_leaderboard_ui_permalink()

# Get project ID
project_id = project.id

def sorted_by_metric(models, test_set, metric):
    models_with_score = [model for model in models if
                         model.metrics[metric][test_set] is not None]
    
    return sorted(models_with_score,
                  key=lambda model: model.metrics[metric][test_set])

# Set styling
dr_dark_blue = '#08233F'
dr_blue = '#1F77B4'
dr_orange = '#FF7F0E'
dr_red = '#BE3C28'

# Function to build histograms
def rebin_df(raw_df, number_of_bins):
    cols = ['bin', 'actual_mean', 'predicted_mean', 'bin_weight']
    new_df = pd.DataFrame(columns=cols)
    current_prediction_total = 0
    current_actual_total = 0
    current_row_total = 0
    x_index = 1
    bin_size = 60 / number_of_bins
    for rowId, data in raw_df.iterrows():
        current_prediction_total += data['predicted'] * data['bin_weight']
        current_actual_total += data['actual'] * data['bin_weight']
        current_row_total += data['bin_weight']

        if ((rowId + 1) % bin_size == 0):
            x_index += 1
            bin_properties = {
                'bin': ((round(rowId + 1) / 60) * number_of_bins),
                'actual_mean': current_actual_total / current_row_total,
                'predicted_mean': current_prediction_total / current_row_total,
                'bin_weight': current_row_total
            }

            new_df = new_df.append(bin_properties, ignore_index=True)
            current_prediction_total = 0
            current_actual_total = 0
            current_row_total = 0
    return new_df

# Function to build lift charts
def matplotlib_lift(bins_df, bin_count, ax):
    grouped = rebin_df(bins_df, bin_count)
    ax.plot(range(1, len(grouped) + 1), grouped['predicted_mean'],
            marker='+', lw=1, color=dr_blue, label='predicted')
    ax.plot(range(1, len(grouped) + 1), grouped['actual_mean'],
            marker='*', lw=1, color=dr_orange, label='actual')
    ax.set_xlim([0, len(grouped) + 1])
    ax.set_facecolor(dr_dark_blue)
    ax.legend(loc='best')
    ax.set_title('Lift chart {} bins'.format(bin_count))
    ax.set_xlabel('Sorted Prediction')
    ax.set_ylabel('Value')
    return grouped

# Create a prediction server
prediction_server = dr.PredictionServer.list()[0]

# Get the top-performing model. Uncomment if this did not execute in the previous cell
# model_top = sorted_by_metric(models, 'crossValidation', metric)[0]

deployment = dr.Deployment.create_from_learning_model(
    model_top.id, label='Predicting Loan Defaults', description='Predicting Loan Defaults',
    default_prediction_server_id=prediction_server.id)
deployment

data_path_scoring = "https://s3.amazonaws.com/datarobot-use-case-datasets/Lending+Club+Dataset+Pred.csv"

scoring_df = pd.read_csv(data_path_scoring, encoding = "ISO-8859-1")

pathfinder_df.rename(columns={'loan_is_bad': 'is_bad'}, inplace=True)

scoring_df

prediction_dataset = project.upload_dataset(scoring_df)
predict_job = model_top.request_predictions(prediction_dataset.id)
prediction_dataset.id

predictions = predict_job.get_result_when_complete()

pd.concat([scoring_df, predictions], axis=1)

# Create forecast CSV
predictions.to_csv("predictions.csv", sep=',', encoding='utf-8',index=False)

# Write Data file to Snowflake
cur.execute("PUT file://predictions.csv @%PREDICTIONS_STAGE;")

cur.execute("COPY INTO PREDICTIONS_STAGE FROM @%PREDICTIONS_TABLE FILE_FORMAT=(type=csv field_delimiter=',', skip_header=1 PURGE=TRUE;")

print("Predictions updated back in Snowflake")