Extract ERS and stress event targets to csv files (completed).

imputation_and_cleaning
Primoz 2022-11-10 09:37:27 +00:00
parent 9199b53ded
commit a668b6e8da
2 changed files with 36 additions and 37 deletions

View File

@ -264,12 +264,14 @@ rule extract_event_information_from_esm:
script:
"../src/features/phone_esm/straw/process_user_event_related_segments.py"
rule create_event_related_segments_file:
rule merge_event_related_segments_files:
input:
ers_files = expand("data/raw/ers/{pid}_ers.csv", pid=config["PIDS"])
ers_files = expand("data/raw/ers/{pid}_ers.csv", pid=config["PIDS"]),
se_files = expand("data/raw/ers/{pid}_stress_event_targets.csv", pid=config["PIDS"])
params:
stage = "merge"
output:
"data/external/straw_events.csv"
"data/external/straw_events.csv",
"data/external/stress_event_targets.csv"
script:
"../src/features/phone_esm/straw/process_user_event_related_segments.py"

View File

@ -23,7 +23,7 @@ def format_timestamp(x):
return tstring
def extract_ers_from_file(esm_df, device_id):
def extract_ers(esm_df, device_id):
pd.set_option("display.max_rows", 20)
pd.set_option("display.max_columns", None)
@ -31,9 +31,7 @@ def extract_ers_from_file(esm_df, device_id):
with open('config.yaml', 'r') as stream:
config = yaml.load(stream, Loader=yaml.FullLoader)
pd.DataFrame().to_csv(snakemake.output[1]) # Create an empty stress event file either way TODO
pd.DataFrame(columns=["label", "intensity"]).to_csv(snakemake.output[1]) # Create an empty stress_events_targets file
esm_preprocessed = clean_up_esm(preprocess_esm(esm_df))
@ -41,7 +39,6 @@ def extract_ers_from_file(esm_df, device_id):
classified = classify_sessions_by_completion_time(esm_preprocessed)
esm_filtered_sessions = classified[classified["session_response"] == 'ema_completed'].reset_index()[['device_id', 'esm_session']]
esm_df = esm_preprocessed.loc[(esm_preprocessed['device_id'].isin(esm_filtered_sessions['device_id'])) & (esm_preprocessed['esm_session'].isin(esm_filtered_sessions['esm_session']))]
targets_method = config["TIME_SEGMENTS"]["TAILORED_EVENTS"]["TARGETS_METHOD"]
if targets_method in ["30_before", "90_before"]: # takes 30-minute peroid before the questionnaire + the duration of the questionnaire
@ -73,14 +70,12 @@ def extract_ers_from_file(esm_df, device_id):
extracted_ers["shift"] = extracted_ers["diffs"].apply(lambda x: format_timestamp(x))
elif targets_method == "stress_event":
# TODO: generiranje ERS datoteke za stress_events
# Get and join required data
extracted_ers = esm_df.groupby(["device_id", "esm_session"])['timestamp'].apply(lambda x: math.ceil((x.max() - x.min()) / 1000)).reset_index().rename(columns={'timestamp': 'session_length'}) # questionnaire end timestamp
session_end_timestamp = esm_df.groupby(['device_id', 'esm_session'])['timestamp'].max().to_frame().rename(columns={'timestamp': 'session_end_timestamp'}) # questionnaire end timestamp
se_time = esm_df[esm_df.questionnaire_id == 90.].set_index(['device_id', 'esm_session'])['esm_user_answer'].to_frame().rename(columns={'esm_user_answer': 'se_time'})
se_duration = esm_df[esm_df.questionnaire_id == 91.].set_index(['device_id', 'esm_session'])['esm_user_answer'].to_frame().rename(columns={'esm_user_answer': 'se_duration'})
se_intensity = esm_df[esm_df.questionnaire_id == 87.].set_index(['device_id', 'esm_session'])['esm_user_answer_numeric'].to_frame().rename(columns={'esm_user_answer_numeric': 'se_intensity'})
se_intensity = esm_df[esm_df.questionnaire_id == 87.].set_index(['device_id', 'esm_session'])['esm_user_answer_numeric'].to_frame().rename(columns={'esm_user_answer_numeric': 'intensity'})
extracted_ers = extracted_ers.join(session_end_timestamp, on=['device_id', 'esm_session'], how='inner') \
.join(se_time, on=['device_id', 'esm_session'], how='inner') \
@ -88,40 +83,30 @@ def extract_ers_from_file(esm_df, device_id):
.join(se_intensity, on=['device_id', 'esm_session'], how='inner')
# Filter sessions that are not useful
extracted_ers = extracted_ers[(extracted_ers.se_time != "0 - Ne spomnim se")]
extracted_ers = extracted_ers[(extracted_ers.se_time != "0 - Ne spomnim se") & (extracted_ers.se_duration != "0 - Ne spomnim se")]
# Transform data into its final form, ready for the extraction
extracted_ers.reset_index(inplace=True)
extracted_ers["label"] = f"straw_event_{targets_method}_" + snakemake.params["pid"] + "_" + extracted_ers.index.astype(str).str.zfill(3)
# Convert to unix timestamp
time_before_event = 90 * 60 # in seconds (10 minutes)
time_before_event = 10 * 60 # in seconds (10 minutes)
extracted_ers['event_timestamp'] = pd.to_datetime(extracted_ers['se_time']).apply(lambda x: x.timestamp() * 1000).astype('int64')
extracted_ers['shift'] = time_before_event
extracted_ers['shift_direction'] = -1
print(extracted_ers[['session_end_timestamp', 'event_timestamp']])
extracted_ers['se_duration'] = \
np.where(
extracted_ers['se_duration'] == "1 - Še vedno traja",
extracted_ers['session_end_timestamp'] - extracted_ers['event_timestamp'],
extracted_ers['se_duration']
)
extracted_ers['se_duration'] = \
np.where(extracted_ers['se_duration'] == "1 - Še vedno traja",
extracted_ers['session_end_timestamp'] - extracted_ers['event_timestamp'],
extracted_ers['se_duration'])
extracted_ers['se_duration'].apply(lambda x: math.ceil(x / 1000) if isinstance(x, int) else (pd.to_datetime(x).hour * 60 + pd.to_datetime(x).minute) * 60) + time_before_event
extracted_ers['se_duration'] = \
extracted_ers['se_duration'].apply(lambda x: math.ceil(x / 1000) if isinstance(x, int) else (pd.to_datetime(x).hour * 60 + pd.to_datetime(x).minute) * 60)
extracted_ers['shift'] = format_timestamp(time_before_event)
extracted_ers['length'] = extracted_ers['se_duration'].apply(lambda x: format_timestamp(x))
sys.exit()
# VV Testiranje različnih povpraševanj za VV
filter_esm = esm_df[(esm_df.esm_type == 7) & ((esm_df.questionnaire_id == 90.) | (esm_df.questionnaire_id == 91.))][['questionnaire_id', 'esm_user_answer', 'esm_session']]
print(filter_esm[filter_esm.esm_user_answer == "1 - Še vedno traja"].shape)
print(filter_esm.shape)
# TODO: generiranje stress_events_targets datoteke (dodaj tudi stolpec s pid) + dodati moraš merge metodo, ki bo združila te datoteke
# TODO: na koncu se mora v čistilni skripti ustrezno odstraniti vse targete in prilepiti nove targete zraven ustreznih segmentov (zna se zgoditi, da bodo overlap)
pd.DataFrame().to_csv(snakemake.output[1])
sys.exit()
extracted_ers[["label", "intensity"]].to_csv(snakemake.output[1], index=False)
else:
raise Exception("Please select correct target method for the event-related segments.")
@ -129,23 +114,35 @@ def extract_ers_from_file(esm_df, device_id):
return extracted_ers[["label", "event_timestamp", "length", "shift", "shift_direction", "device_id"]]
# Actual code execution
if snakemake.params["stage"] == "extract":
esm_df = pd.read_csv(input_data_files['esm_raw_input'])
with open(input_data_files['pid_file'], 'r') as stream:
pid_file = yaml.load(stream, Loader=yaml.FullLoader)
extracted_ers = extract_ers_from_file(esm_df, pid_file["PHONE"]["DEVICE_IDS"][0])
extracted_ers = extract_ers(esm_df, pid_file["PHONE"]["DEVICE_IDS"][0])
extracted_ers.to_csv(snakemake.output[0], index=False)
elif snakemake.params["stage"] == "merge":
elif snakemake.params["stage"] == "merge":
input_data_files = dict(snakemake.input)
straw_events = pd.DataFrame(columns=["label", "event_timestamp", "length", "shift", "shift_direction", "device_id"])
stress_events_targets = pd.DataFrame(columns=["label", "intensity"])
for input_file in input_data_files["ers_files"]:
ers_df = pd.read_csv(input_file)
straw_events = pd.concat([straw_events, ers_df], axis=0, ignore_index=True)
straw_events.to_csv(snakemake.output[0], index=False)
for input_file in input_data_files["se_files"]:
se_df = pd.read_csv(input_file)
stress_events_targets = pd.concat([stress_events_targets, se_df], axis=0, ignore_index=True)
stress_events_targets.to_csv(snakemake.output[1], index=False)