Hands-On Deep Learning for IoT

Making predictions

Now we can use our created features to fit RandomForestRegressor to our data and see whether we can improve on the previous scores:

rf = RandomForestRegressor() # first we instantiate RandomForestRegressor from scikit-learnrf.fit(X_train, y_train) # train the regressor model with traing set   preds = rf.predict(X_test) # making predictin on unseen observation scores = mean_absolute_error(preds, y_test) # Computing MAEprint('Mean Abs Error: {:.2f}'.format(scores))high_imp_feats = utils.feature_importances(X, reg, feats=10) # Printing feature importance

The preceding code block should produce the following output showing the baseline MAE values and statistics about the engine recording cycles:

Mean Abs Error: 31.04 1: LAST(recordings.cycles.LAST(recordings.sensor_measurement_4)) ...