junos
/
rapids
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

Add day epochs to battery metrics and fix some of them

replace/8899df1725485ee49bd919daba6d6aa5f7a6bd2c
Meng Li 2019-11-25 12:53:32 -05:00
parent d8421575ba
commit 99d387a7a4
4 changed files with 115 additions and 51 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
Snakefile
View File

@ -23,7 +23,9 @@ rule all:
segment = config["BLUETOOTH"]["DAY_SEGMENTS"]), segment = config["BLUETOOTH"]["DAY_SEGMENTS"]),
expand("data/processed/{pid}/google_activity_recognition_{segment}.csv",pid=config["PIDS"], expand("data/processed/{pid}/google_activity_recognition_{segment}.csv",pid=config["PIDS"],
segment = config["GOOGLE_ACTIVITY_RECOGNITION"]["DAY_SEGMENTS"]), segment = config["GOOGLE_ACTIVITY_RECOGNITION"]["DAY_SEGMENTS"]),
expand("data/processed/{pid}/battery_daily.csv", pid=config["PIDS"]), expand("data/processed/{pid}/battery_{day_segment}.csv",
pid = config["PIDS"],
day_segment = config["BATTERY"]["DAY_SEGMENTS"]),
# Reports # Reports
expand("reports/figures/{pid}/{sensor}_heatmap_rows.html", pid=config["PIDS"], sensor=config["SENSORS"]), expand("reports/figures/{pid}/{sensor}_heatmap_rows.html", pid=config["PIDS"], sensor=config["SENSORS"]),
expand("reports/figures/{pid}/compliance_heatmap.html", pid=config["PIDS"], sensor=config["SENSORS"]), expand("reports/figures/{pid}/compliance_heatmap.html", pid=config["PIDS"], sensor=config["SENSORS"]),

5
config.yaml
View File

@ -57,3 +57,8 @@ BLUETOOTH:
GOOGLE_ACTIVITY_RECOGNITION: GOOGLE_ACTIVITY_RECOGNITION:
DAY_SEGMENTS: *day_segments DAY_SEGMENTS: *day_segments
METRICS: ['count','most_common_activity','number_unique_activities','activity_change_count'] METRICS: ['count','most_common_activity','number_unique_activities','activity_change_count']
BATTERY:
DAY_SEGMENTS: *day_segments
METRICS: ["countdischarge", "sumdurationdischarge", "countcharge", "sumdurationcharge", "avgconsumptionrate", "maxconsumptionrate"]

5
rules/features.snakefile
View File

@ -66,7 +66,10 @@ rule activity_metrics:
rule battery_metrics: rule battery_metrics:
input: input:
"data/processed/{pid}/battery_deltas.csv" "data/processed/{pid}/battery_deltas.csv"
params:
day_segment = "{day_segment}",
metrics = config["BATTERY"]["METRICS"]
output: output:
"data/processed/{pid}/battery_daily.csv" "data/processed/{pid}/battery_{day_segment}.csv"
script: script:
"../src/features/battery_metrics.py" "../src/features/battery_metrics.py"

152
src/features/battery_metrics.py
View File

@ -1,60 +1,114 @@
import pandas as pd import pandas as pd
import datetime from datetime import datetime, timedelta, time
def truncateTime(df, segment_column, new_day_segment, datetime_column, date_column, new_time):
df.loc[:, segment_column] = new_day_segment
df.loc[:, datetime_column] = df[date_column].apply(lambda date: datetime.combine(date, new_time))
return df
def computeTruncatedBatteryTimeDifferences(df):
df["truncated_time_diff"] = df["local_end_date_time"] - df["local_start_date_time"]
df["truncated_time_diff"] = df["truncated_time_diff"].apply(lambda time: time.total_seconds()/3600)
df["battery_diff"] = df["battery_diff"] * (df["truncated_time_diff"] / df["time_diff"])
del df["time_diff"]
df.rename(columns={"truncated_time_diff": "time_diff"}, inplace=True)
return df
def splitOvernightEpisodes(battery_data):
overnight = battery_data[(battery_data["local_start_date"] + timedelta(days=1)) == battery_data["local_end_date"]]
not_overnight = battery_data[battery_data["local_start_date"] == battery_data["local_end_date"]]
if not overnight.empty:
today = overnight[["battery_diff", "time_diff", "local_start_date_time", "local_start_date", "local_start_day_segment"]].copy()
tomorrow = overnight[["battery_diff", "time_diff", "local_end_date_time", "local_end_date", "local_end_day_segment"]].copy()
# truncate the end time of all overnight periods to midnight
today = truncateTime(today, "local_end_day_segment", "evening", "local_end_date_time", "local_start_date", time(23,59,59))
today["local_end_date"] = overnight["local_start_date"]
# set the start time of all periods after midnight to midnight
tomorrow = truncateTime(tomorrow, "local_start_day_segment", "night", "local_start_date_time", "local_end_date", time(0,0,0))
tomorrow["local_start_date"] = overnight["local_end_date"]
overnight = pd.concat([today, tomorrow], axis=0, sort=False)
# calculate new battery_diff and time_diff for split overnight periods
overnight = computeTruncatedBatteryTimeDifferences(overnight)
return pd.concat([not_overnight, overnight], axis=0, sort=False)
def splitMultiSegmentEpisodes(battery_data, day_segment):
# extract episodes that start and end at the same epochs
exact_segments = battery_data.query("local_start_day_segment == local_end_day_segment and local_start_day_segment == @day_segment").copy()
# extract episodes that start and end at different epochs
across_segments = battery_data.query("local_start_day_segment != local_end_day_segment").copy()
# 1) if start time is in current day_segment
start_segment = across_segments[across_segments["local_start_day_segment"] == day_segment].copy()
if not start_segment.empty:
start_segment = truncateTime(start_segment, "local_end_day_segment", day_segment, "local_end_date_time", "local_end_date", time(EPOCH_TIMES[day_segment][1],59,59))
# 2) if end time is in current day_segment
end_segment = across_segments[across_segments["local_end_day_segment"] == day_segment].copy()
if not end_segment.empty:
end_segment = truncateTime(end_segment, "local_start_day_segment", day_segment, "local_start_date_time", "local_start_date", time(EPOCH_TIMES[day_segment][0],0,0))
# 3) if current episode comtains day_segment
across_segments.loc[:,"start_segment"] = across_segments["local_start_day_segment"].apply(lambda seg: SEGMENT[seg])
across_segments.loc[:,"end_segment"] = across_segments["local_end_day_segment"].apply(lambda seg: SEGMENT[seg])
day_segment_num = SEGMENT[day_segment]
within_segments = across_segments.query("start_segment < @day_segment_num and end_segment > @day_segment_num")
del across_segments["start_segment"], across_segments["end_segment"]
del within_segments["start_segment"], within_segments["end_segment"]
if not within_segments.empty:
within_segments = truncateTime(within_segments, "local_start_day_segment", day_segment, "local_start_date_time", "local_start_date", time(EPOCH_TIMES[day_segment][0],0,0))
within_segments = truncateTime(within_segments, "local_end_day_segment", day_segment, "local_end_date_time", "local_end_date", time(EPOCH_TIMES[day_segment][1],59,59))
across_segments = pd.concat([start_segment, end_segment, within_segments], axis=0, sort=False)
if not across_segments.empty:
accross_segments = computeTruncatedBatteryTimeDifferences(across_segments)
return pd.concat([exact_segments, across_segments], axis=0, sort=False)
battery_data = pd.read_csv(snakemake.input[0]) battery_data = pd.read_csv(snakemake.input[0], parse_dates=["local_start_date_time", "local_end_date_time", "local_start_date", "local_end_date"])
day_segment = snakemake.params["day_segment"]
metrics = snakemake.params["metrics"]
SEGMENT = {"night": 0, "morning": 1, "afternoon": 2, "evening": 3}
EPOCH_TIMES = {"night": [0,5], "morning": [6,11], "afternoon": [12,17], "evening": [18,23]}
if battery_data.empty: if battery_data.empty:
battery_features = pd.DataFrame(columns=["battery_diff", "time_diff", "battery_decrease_times","battery_consumption_rate", "local_date"]) battery_features = pd.DataFrame(columns=["local_date"] + ["battery_" + day_segment + "_" + x for x in metrics])
else: else:
for col in ["local_start_date_time", "local_end_date_time", "local_start_date", "local_end_date"]: battery_data = splitOvernightEpisodes(battery_data)
battery_data[col] = pd.to_datetime(battery_data[col])
# split the row into 2 rows when local_start_date + 1 = local_end_date if day_segment != "daily":
battery_data_overnight = battery_data[battery_data["local_start_date"] + datetime.timedelta(days=1) == battery_data["local_end_date"]] battery_data = splitMultiSegmentEpisodes(battery_data, day_segment)
if not battery_data_overnight.empty:
battery_data_overnight_first, battery_data_overnight_second = pd.DataFrame(columns=battery_data.columns), pd.DataFrame(columns=battery_data.columns)
battery_data_overnight_first["total_battery_diff"], battery_data_overnight_second["total_battery_diff"] = battery_data_overnight["battery_diff"], battery_data_overnight["battery_diff"]
battery_data_overnight_first["total_time_diff"], battery_data_overnight_second["total_time_diff"] = battery_data_overnight["time_diff"], battery_data_overnight["time_diff"]
# let start = start OR end = end, then fill the left col with the start+23:59:59.
battery_data_overnight_first["local_start_date_time"] = battery_data_overnight["local_start_date_time"]
battery_data_overnight_first["local_end_date_time"] = battery_data_overnight["local_start_date"].apply(lambda x: datetime.datetime.combine(x, datetime.time(23,59,59)))
battery_data_overnight_first["local_start_date"] = battery_data_overnight["local_start_date"]
battery_data_overnight_second["local_end_date_time"] = battery_data_overnight["local_end_date_time"]
battery_data_overnight_second["local_start_date_time"] = battery_data_overnight["local_start_date"].apply(lambda x: datetime.datetime.combine(x, datetime.time(23,59,59)))
battery_data_overnight_second["local_start_date"] = battery_data_overnight["local_end_date"]
battery_data_overnight = pd.concat([battery_data_overnight_first, battery_data_overnight_second])
# calculate battery_diff and time_diff
battery_data_overnight["time_diff"] = (battery_data_overnight["local_end_date_time"]-battery_data_overnight["local_start_date_time"]).apply(lambda x: x.total_seconds()/3600)
battery_data_overnight["battery_diff"] = battery_data_overnight["total_battery_diff"]*(battery_data_overnight["time_diff"]/battery_data_overnight["total_time_diff"])
del battery_data_overnight["total_battery_diff"], battery_data_overnight["total_time_diff"]
# filter out the rows when local_start_date + 1 < local_end_date battery_data["battery_consumption_rate"] = battery_data["battery_diff"] / battery_data["time_diff"]
battery_data = battery_data[battery_data["local_start_date"] == battery_data["local_end_date"]]
# combine # for battery_data_discharge:
battery_data = pd.concat([battery_data, battery_data_overnight]) battery_data_discharge = battery_data[battery_data["battery_diff"] > 0]
battery_discharge_features = pd.DataFrame()
# split into decrease table and charge table if "countdischarge" in metrics:
battery_data_decrease = battery_data[battery_data["battery_diff"] > 0] battery_discharge_features["battery_"+day_segment+"_countdischarge"] = battery_data_discharge.groupby(["local_start_date"])["local_start_date"].count()
battery_data_charge = battery_data[battery_data["battery_diff"] <= 0] if "sumdurationdischarge" in metrics:
battery_discharge_features["battery_"+day_segment+"_sumdurationdischarge"] = battery_data_discharge.groupby(["local_start_date"])["time_diff"].sum()
# for battery_data_decrease: if "avgconsumptionrate" in metrics:
battery_decrease_count = battery_data_decrease.groupby(["local_start_date"])["local_start_date"].count() battery_discharge_features["battery_"+day_segment+"_avgconsumptionrate"] = battery_data_discharge.groupby(["local_start_date"])["battery_consumption_rate"].mean()
battery_data_decrease = battery_data_decrease.groupby(["local_start_date"]).sum() if "maxconsumptionrate" in metrics:
battery_data_decrease["battery_decrease_count"] = battery_decrease_count battery_discharge_features["battery_"+day_segment+"_maxconsumptionrate"] = battery_data_discharge.groupby(["local_start_date"])["battery_consumption_rate"].max()
battery_data_decrease["battery_decrease_duration"] = battery_data_decrease["time_diff"]
battery_data_decrease["battery_consumption_rate"] = battery_data_decrease["battery_diff"]/battery_data_decrease["time_diff"]
del battery_data_decrease["battery_diff"], battery_data_decrease["time_diff"]
# for battery_data_charge: # for battery_data_charge:
battery_charge_count = battery_data_charge.groupby(["local_start_date"])["local_start_date"].count() battery_data_charge = battery_data[battery_data["battery_diff"] <= 0]
battery_data_charge = battery_data_charge.groupby(["local_start_date"]).sum() battery_charge_features = pd.DataFrame()
battery_data_charge["battery_charge_count"] = battery_charge_count if "countcharge" in metrics:
battery_data_charge["battery_charge_duration"] = battery_data_charge["time_diff"] battery_charge_features["battery_"+day_segment+"_countcharge"] = battery_data_charge.groupby(["local_start_date"])["local_start_date"].count()
del battery_data_charge["battery_diff"], battery_data_charge["time_diff"] if "sumdurationcharge" in metrics:
battery_charge_features["battery_"+day_segment+"_sumdurationcharge"] = battery_data_charge.groupby(["local_start_date"])["time_diff"].sum()
# combine decrease features and charge features # combine discharge features and charge features; fill the missing values with ZERO
battery_features = pd.concat([battery_data_decrease, battery_data_charge], axis=1, sort=True) battery_features = pd.concat([battery_discharge_features, battery_charge_features], axis=1, sort=True).fillna(0)
battery_features["local_date"] = battery_features.index battery_features.index.rename("local_date", inplace=True)
battery_features.reset_index(inplace=True, drop=True) battery_features.to_csv(snakemake.output[0], index=True)
battery_features.to_csv(snakemake.output[0], index=False)