tsfresh_robot_failures_feature_extraction_and_selection.py

1from tsfresh import extract_features, select_features
2from tsfresh.utilities.dataframe_functions import impute
3from tsfresh.feature_extraction import ComprehensiveFCParameters
4from tsfresh.examples.robot_execution_failures import download_robot_execution_failures, \
5    load_robot_execution_failures
6import pandas as pd
7
8# 1. Download and load the data
9download_robot_execution_failures()
10timeseries, y = load_robot_execution_failures()
11
12# 2. Extract features
13# We use the ComprehensiveFCParameters which includes all available features
14settings = ComprehensiveFCParameters()
15
16extracted_features = extract_features(timeseries, column_id="id", column_sort="time",
17                                      default_fc_parameters=settings,
18                                      # we impute meaningful values for NaN features
19                                      impute_function=impute)
20
21# 3. Select relevant features
22# This will perform a statistical hypothesis test to see which features 
23# are relevant for the target y
24features_filtered = select_features(extracted_features, y)
25
26# Show the result
27print(features_filtered.head())