python_code
stringlengths 0
4.04M
| repo_name
stringlengths 7
58
| file_path
stringlengths 5
147
|
|---|---|---|
#!/usr/bin/env python3
# Copyright (c) 2021-present, Facebook, Inc.
# All rights reserved.
#
# This source code is licensed under the license found in the
# LICENSE file in the root directory of this source tree.
import numpy as np
import pandas as pd
from hdx.api.configuration import Configuration
from hdx.data.dataset import Dataset
import shutil
from glob import glob
import os
from covid19_spread.data.usa.process_cases import get_index
import re
SCRIPT_DIR = os.path.dirname(os.path.realpath(__file__))
def main():
Configuration.create(
hdx_site="prod", user_agent="A_Quick_Example", hdx_read_only=True
)
dataset = Dataset.read_from_hdx("movement-range-maps")
resources = dataset.get_resources()
resource = [
x
for x in resources
if re.match(".*/movement-range-data-\d{4}-\d{2}-\d{2}\.zip", x["url"])
]
assert len(resource) == 1
resource = resource[0]
url, path = resource.download()
if os.path.exists(f"{SCRIPT_DIR}/fb_mobility"):
shutil.rmtree(f"{SCRIPT_DIR}/fb_mobility")
shutil.unpack_archive(path, f"{SCRIPT_DIR}/fb_mobility", "zip")
fips_map = get_index()
fips_map["location"] = fips_map["name"] + ", " + fips_map["subregion1_name"]
cols = [
"date",
"region",
"all_day_bing_tiles_visited_relative_change",
"all_day_ratio_single_tile_users",
]
def get_county_mobility_fb(fin):
df_mobility_global = pd.read_csv(
fin, parse_dates=["ds"], delimiter="\t", dtype={"polygon_id": str}
)
df_mobility_usa = df_mobility_global.query("country == 'USA'")
return df_mobility_usa
# fin = sys.argv[1] if len(sys.argv) == 2 else None
txt_files = glob(f"{SCRIPT_DIR}/fb_mobility/movement-range*.txt")
assert len(txt_files) == 1
fin = txt_files[0]
df = get_county_mobility_fb(fin)
df = df.rename(columns={"ds": "date", "polygon_id": "region"})
df = df.merge(fips_map, left_on="region", right_on="fips")[
list(df.columns) + ["location"]
]
df = df.drop(columns="region").rename(columns={"location": "region"})
def zscore(df):
# z-scores
df = (df.values - df.mean(skipna=True)) / df.std(skipna=True)
return df
def process_df(df, cols):
df = df[cols].copy()
regions = []
for (name, _df) in df.groupby("region"):
_df = _df.sort_values(by="date")
_df = _df.drop_duplicates(subset="date")
dates = _df["date"].to_list()
assert len(dates) == len(np.unique(dates)), _df
_df = _df.loc[:, ~_df.columns.duplicated()]
_df = _df.drop(columns=["region", "date"]).transpose()
# take 7 day average
_df = _df.rolling(7, min_periods=1, axis=1).mean()
# convert relative change into absolute numbers
_df.loc["all_day_bing_tiles_visited_relative_change"] += 1
_df["region"] = [name] * len(_df)
_df.columns = list(map(lambda x: x.strftime("%Y-%m-%d"), dates)) + [
"region"
]
regions.append(_df.reset_index())
df = pd.concat(regions, axis=0, ignore_index=True)
cols = ["region"] + [x for x in df.columns if x != "region"]
df = df[cols]
df = df.rename(columns={"index": "type"})
return df
county = process_df(df, cols)
state = df.copy()
state["region"] = state["region"].apply(lambda x: x.split(", ")[-1])
state = state.groupby(["region", "date"]).mean().reset_index()
state = process_df(state, cols)
county = county.fillna(0)
state = state.fillna(0)
county.round(4).to_csv(f"{SCRIPT_DIR}/mobility_features_county_fb.csv", index=False)
state.round(4).to_csv(f"{SCRIPT_DIR}/mobility_features_state_fb.csv", index=False)
if __name__ == "__main__":
main()
| covid19_spread-main | covid19_spread/data/usa/fb/process_mobility.py |
#!/usr/bin/env python3
# Copyright (c) 2021-present, Facebook, Inc.
# All rights reserved.
#
# This source code is licensed under the license found in the
# LICENSE file in the root directory of this source tree.
from .process_mobility import main
def prepare():
main()
| covid19_spread-main | covid19_spread/data/usa/fb/__init__.py |
#!/usr/bin/env python3
# Copyright (c) 2021-present, Facebook, Inc.
# All rights reserved.
#
# This source code is licensed under the license found in the
# LICENSE file in the root directory of this source tree.
from .process_testing import main
def prepare():
main()
| covid19_spread-main | covid19_spread/data/usa/testing/__init__.py |
#!/usr/bin/env python3
# Copyright (c) 2021-present, Facebook, Inc.
# All rights reserved.
#
# This source code is licensed under the license found in the
# LICENSE file in the root directory of this source tree.
import pandas as pd
from datetime import datetime
import os
from covid19_spread.data.usa.process_cases import get_index
SCRIPT_DIR = os.path.dirname(os.path.realpath(__file__))
def main():
df = pd.read_csv(
"https://beta.healthdata.gov/api/views/j8mb-icvb/rows.csv?accessType=DOWNLOAD",
parse_dates=["date"],
)
df_piv = df.pivot(
columns=["overall_outcome"],
values="total_results_reported",
index=["state", "date"],
)
df_piv = df_piv.fillna(0).groupby(level=0).cummax()
index = get_index()
states = index.drop_duplicates("subregion1_name")
with_index = df_piv.reset_index().merge(
states, left_on="state", right_on="subregion1_code"
)
df = with_index[
["subregion1_name", "Negative", "Positive", "Inconclusive", "date"]
].set_index("date")
df = df.rename(columns={"subregion1_name": "state_name"})
df["Total"] = df["Positive"] + df["Negative"] + df["Inconclusive"]
def zscore(df):
df.iloc[:, 0:] = (
df.iloc[:, 0:].values
- df.iloc[:, 0:].mean(axis=1, skipna=True).values[:, None]
) / df.iloc[:, 0:].std(axis=1, skipna=True).values[:, None]
df = df.fillna(0)
return df
def zero_one(df):
df = df.fillna(0)
df = df.div(df.max(axis=1), axis=0)
# df = df / df.max()
df = df.fillna(0)
return df
def fmt_features(pivot, key, func_smooth, func_normalize):
df = pivot.transpose()
df = func_smooth(df)
if func_normalize is not None:
df = func_normalize(df)
df = df.fillna(0)
df.index.set_names("region", inplace=True)
df["type"] = f"testing_{key}"
merge = df.merge(index, left_index=True, right_on="subregion1_name")
merge.index = merge["name"] + ", " + merge["subregion1_name"]
return df, merge[df.columns]
def _diff(df):
return df.diff(axis=1).rolling(7, axis=1, min_periods=1).mean()
state_r, county_r = fmt_features(
df.pivot(columns="state_name", values=["Positive", "Total"]),
"ratio",
lambda _df: (_diff(_df.loc["Positive"]) / _diff(_df.loc["Total"])),
None,
)
state_t, county_t = fmt_features(
df.pivot(columns="state_name", values="Total"), "Total", _diff, zero_one,
)
def write_features(df, res, fout):
df = df[["type"] + [c for c in df.columns if isinstance(c, datetime)]]
df.columns = [
str(x.date()) if isinstance(x, datetime) else x for x in df.columns
]
df.round(3).to_csv(
f"{SCRIPT_DIR}/{fout}_features_{res}.csv", index_label="region"
)
write_features(state_t, "state", "total")
write_features(state_r, "state", "ratio")
write_features(county_t, "county", "total")
write_features(county_r, "county", "ratio")
if __name__ == "__main__":
main()
| covid19_spread-main | covid19_spread/data/usa/testing/process_testing.py |
#!/usr/bin/env python3
# Copyright (c) 2021-present, Facebook, Inc.
# All rights reserved.
#
# This source code is licensed under the license found in the
# LICENSE file in the root directory of this source tree.
import os
import pandas as pd
import sys
from datetime import timedelta
from delphi_epidata import Epidata
import covidcast
# Fetch data
SCRIPT_DIR = os.path.dirname(os.path.realpath(__file__))
def main(geo_value, source, signal):
# grab start and end date from metadata
df = covidcast.metadata().drop(columns=["time_type", "min_lag", "max_lag"])
df.min_time = pd.to_datetime(df.min_time)
df.max_time = pd.to_datetime(df.max_time)
df = df.query(
f"data_source == '{source}' and signal == '{signal}' and geo_type == '{geo_value}'"
)
assert len(df) == 1
base_date = df.iloc[0].min_time - timedelta(1)
end_date = df.iloc[0].max_time
dfs = []
current_date = base_date
while current_date < end_date:
current_date = current_date + timedelta(1)
date_str = current_date.strftime("%Y%m%d")
os.makedirs(os.path.join(SCRIPT_DIR, geo_value, source), exist_ok=True)
fout = f"{SCRIPT_DIR}/{geo_value}/{source}/{signal}-{date_str}.csv"
# d/l only if we don't have the file already
if os.path.exists(fout):
dfs.append(pd.read_csv(fout))
continue
for _ in range(3):
res = Epidata.covidcast(source, signal, "day", geo_value, [date_str], "*")
print(date_str, res["result"], res["message"])
if res["result"] == 1:
break
if res["result"] != 1:
# response may be non-zero if there aren't enough respondants
# See: https://github.com/cmu-delphi/delphi-epidata/issues/613#event-4962274038
print(f"Skipping {source}/{signal} for {date_str}")
continue
df = pd.DataFrame(res["epidata"])
df.rename(
columns={
"geo_value": geo_value,
"time_value": "date",
"value": signal,
"direction": f"{signal}_direction",
"stderr": f"{signal}_stderr",
"sample_size": f"{signal}_sample_size",
},
inplace=True,
)
df.to_csv(fout, index=False)
dfs.append(df)
pd.concat(dfs).to_csv(f"{SCRIPT_DIR}/{geo_value}/{source}/{signal}.csv")
SIGNALS = [
("fb-survey", "smoothed_hh_cmnty_cli"),
("fb-survey", "smoothed_wcli"),
("doctor-visits", "smoothed_adj_cli"),
("fb-survey", "smoothed_wcovid_vaccinated_or_accept"),
("fb-survey", "smoothed_wearing_mask"),
("fb-survey", "smoothed_wearing_mask_7d"),
("fb-survey", "smoothed_wothers_masked"),
("fb-survey", "smoothed_wcovid_vaccinated_or_accept"),
]
if __name__ == "__main__":
main(sys.argv[1], *sys.argv[2].split("/"))
| covid19_spread-main | covid19_spread/data/usa/symptom_survey/fetch.py |
#!/usr/bin/env python3
# Copyright (c) 2021-present, Facebook, Inc.
# All rights reserved.
#
# This source code is licensed under the license found in the
# LICENSE file in the root directory of this source tree.
import pandas as pd
import argparse
from datetime import datetime
import os
from covid19_spread.data.usa.process_cases import get_index
SCRIPT_DIR = os.path.dirname(os.path.realpath(__file__))
def get_df(source, signal, resolution):
df = pd.read_csv(
f"{SCRIPT_DIR}/{resolution}/{source}/{signal}.csv", parse_dates=["date"]
)
df.dropna(axis=0, subset=["date"], inplace=True)
index = get_index()
state_index = index.drop_duplicates("subregion1_code")
if "state" in df.columns:
df["state"] = df["state"].str.upper()
merged = df.merge(state_index, left_on="state", right_on="subregion1_code")
df = merged[["subregion1_name", "date", signal]].rename(
columns={"subregion1_name": "loc"}
)
else:
df["county"] = df["county"].astype(str).str.zfill(5)
merged = df.merge(index, left_on="county", right_on="fips")
merged["loc"] = merged["name"] + ", " + merged["subregion1_name"]
df = merged[["loc", "date", signal]]
df = df.pivot(index="date", columns="loc", values=signal).copy()
# Fill in NaNs
df.iloc[0] = 0
df = df.fillna(0)
# Normalize
df = df.transpose() / 100
df["type"] = f"{source}_{signal}_{resolution}"
return df
def main(signal, resolution):
source, signal = signal.split("/")
df = get_df(source, signal, resolution)
if resolution == "county":
# Fill in missing counties with zeros
cases = pd.read_csv(
f"{SCRIPT_DIR}/../data_cases.csv", index_col="region"
).index.to_frame()
cases["state"] = [x.split(", ")[-1] for x in cases.index]
cases = cases.drop(columns="region")
idx = pd.MultiIndex.from_product([cases.index, df["type"].unique()])
type_ = df["type"].iloc[0]
df = df.reset_index().set_index(["loc", "type"]).reindex(idx).fillna(0)
df2 = get_df(source, signal, "state")
df2 = df2.merge(cases[["state"]], left_index=True, right_on="state")[
df2.columns
]
df = pd.concat([df, df2.set_index("type", append=True)])
df = df[[c for c in df.columns if isinstance(c, datetime)]]
df.columns = [str(x.date()) if isinstance(x, datetime) else x for x in df.columns]
df.round(3).to_csv(
f"{SCRIPT_DIR}/{source}_{signal}-{resolution}.csv",
index_label=["region", "type"],
)
if __name__ == "__main__":
parser = argparse.ArgumentParser()
parser.add_argument("-signal", default="smoothed_hh_cmnty_cli")
parser.add_argument("-resolution", choices=["state", "county"], default="county")
opt = parser.parse_args()
main(opt.signal, opt.resolution)
| covid19_spread-main | covid19_spread/data/usa/symptom_survey/process_symptom_survey.py |
#!/usr/bin/env python3
# Copyright (c) 2021-present, Facebook, Inc.
# All rights reserved.
#
# This source code is licensed under the license found in the
# LICENSE file in the root directory of this source tree.
from covid19_spread.data.usa.symptom_survey.fetch import main as fetch, SIGNALS
from covid19_spread.data.usa.symptom_survey.process_symptom_survey import (
main as process,
)
import os
import pandas
SCRIPT_DIR = os.path.dirname(os.path.realpath(__file__))
def concat_mask_data(resolution):
df1 = pandas.read_csv(
os.path.join(SCRIPT_DIR, resolution, "fb-survey", "smoothed_wearing_mask.csv")
)
df2 = pandas.read_csv(
os.path.join(
SCRIPT_DIR, resolution, "fb-survey", "smoothed_wearing_mask_7d.csv"
)
)
df2.columns = [c.replace("_7d", "") for c in df2.columns]
df = pandas.concat([df1, df2])
df.columns = [
c.replace("smoothed_wearing_mask", "smoothed_wearing_mask_all")
for c in df.columns
]
df = df.sort_values(by="signal", ascending=False)
df["signal"] = "smoothed_wearing_mask_all"
df = df.drop_duplicates([resolution, "date"])
df.to_csv(
os.path.join(
SCRIPT_DIR, resolution, "fb-survey", "smoothed_wearing_mask_all.csv"
),
index=False,
)
def prepare():
for source, signal in SIGNALS:
fetch("state", source, signal)
fetch("county", source, signal)
concat_mask_data("county")
concat_mask_data("state")
for source, signal in SIGNALS:
if "wearing_mask" in signal:
# Skip these since we end up concatenating the wearing_mask and wearing_mask_7d features
continue
process(f"{source}/{signal}", "state")
process(f"{source}/{signal}", "county")
process(f"fb-survey/smoothed_wearing_mask_all", "county")
process(f"fb-survey/smoothed_wearing_mask_all", "state")
if __name__ == "__main__":
prepare()
| covid19_spread-main | covid19_spread/data/usa/symptom_survey/__init__.py |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.