n_caseids = filtered_event_df["caseid"].unique().shape[0]
filter_by_conforming_caseids = filtered_event_df["caseid"].isin(
conforming_caseids)
filtered_train_event_df = filtered_event_df.loc[
~filter_by_conforming_caseids, :]
n_caseids_train = filtered_train_event_df["caseid"].unique().shape[0]
logger.info(
f"Filtered train event df shape: {filtered_train_event_df.shape}")
logger.info(
f'Fitting with {n_caseids_train}/{n_caseids} non-conforming cases')
train_X, train_lengths = event_df_to_hmm_format(filtered_train_event_df)
logger.info(f"Starting training...")
start_fit = time.time()
tracker = hmmconf.ConformanceTracker(
hmm, max_n_case=EXPERIMENT_CONFIGS[MAX_N_CASE])
tracker.hmm.fit(train_X, train_lengths)
took_fit = time.time() - start_fit
time_dict[TIME_FIT] = took_fit
time_dict[TIME_N_TRAIN_CASES] = n_caseids_train
time_dict[TIME_N_TRAIN_EVENTS] = filtered_train_event_df.shape[0]
info_msg = f"Training using {n_caseids_train} cases took: {took_fit:.3f}s"
info_msg += f" ({took_fit / 60:.0f} mins {took_fit % 60:.0f} secs)"
logger.info(info_msg)
took_all = time.time() - start_all
logger.info(f"Took: {took_all / 60:.0f} mins {took_all % 60:.0f} secs")
time_dict[TIME_ALL] = took_all
time_fp = os.path.join(results_dir, "time_results.csv")
time_df = pd.DataFrame(time_dict, index=[0])
def make_conformance_tracker(hmm):
return hmmconf.ConformanceTracker(hmm, max_n_case=100000)