-
Notifications
You must be signed in to change notification settings - Fork 0
/
preprocess_data.py
202 lines (168 loc) · 8.12 KB
/
preprocess_data.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
import numpy as np
import pandas as pd
from datetime import datetime, timezone, timedelta
from dateutil.parser import parse
import matplotlib.pyplot as plt
import math
import os
import pickle
from absl import app
from absl import flags
FLAGS = flags.FLAGS
flags.DEFINE_string('start_date', '', 'Start date for importing data. Format is YYYY-MM-dd. If not specified, '
'will default to creating the entire table if the table does not already exist, and doing '
'incremental updates if the table already exists.')
DEFAULT_START_DATE = datetime(2019, 1, 1) # Set some long time in the past before infection.
def is_date(string, fuzzy=False):
"""Return whether the string can be interpreted as a date.
Args:
string: String to check for date.
Args:
fuzzy: Ignore unknown tokens in string if True.
"""
try:
parse(string, fuzzy=fuzzy)
return True
except ValueError:
return False
def get_county(string):
"""Parse out a US county from a combined string containing county, state, and country.
Args:
string: The full string in format "county, state, country".
Returns:
The county name if it exists, otherwise empty string.
"""
str_array = string.split(',')
if len(str_array) < 3:
return ''
else:
return str_array[0]
def write_us_county_data(us_confirmed_df, us_deaths_df, county_census_df, filename, start_date=''):
column_names = ['FIPS', 'County', 'Province_State', 'Country_Region', 'Date', 'Cases', 'Deaths', 'Population']
if os.path.exists(filename):
with open(filename, 'rb') as f:
us_combined_df = pickle.load(f)
else:
us_combined_df = pd.DataFrame(columns=column_names)
if not start_date:
if len(us_combined_df):
start_date = us_combined_df['Date'].max() + timedelta(0, 0, 1)
else:
start_date = DEFAULT_START_DATE
else:
start_date = datetime.strptime(start_date, '%Y-%m-%d')
# Use date conversions to extract correct dates from column names in COVID dataset
date_cols = [x for x in list(us_confirmed_df) if is_date(x)]
dates = [datetime.strptime(x, '%m/%d/%y') for x in date_cols]
dates = [x for x in dates if x >= start_date]
date_cols = [x for x in date_cols if datetime.strptime(x, '%m/%d/%y') >= start_date]
# Append rows that have confirmed cases, deaths, and populations included.
for index, row in us_confirmed_df.iterrows():
fips = row['FIPS']
county = get_county(row['Combined_Key'])
if math.isnan(fips):
print('skipping county', county, fips, population)
continue
population = county_census_df[
(county_census_df.STATE == int(fips / 1000))
& (county_census_df.COUNTY == int(fips % 1000))]['POPESTIMATE2019']
if len(population) != 1:
print('skipping county', county, fips, population)
continue
population = population.to_numpy()[0]
for (date_col, date) in zip(date_cols, dates):
confirmed = row[date_col]
if confirmed == 0:
continue
if date_col in us_deaths_df:
deaths = us_deaths_df[us_deaths_df.FIPS == row['FIPS']][date_col].to_numpy()[0]
else:
deaths = 0
values = [fips, county, row['Province_State'], row['Country_Region'], date, confirmed, deaths, population]
df_length = len(us_combined_df)
us_combined_df.loc[df_length] = values
if index % 100 == 0:
print('processed {} out of {}'.format(index, len(us_confirmed_df)))
us_combined_df = us_combined_df.drop_duplicates(['Date', 'FIPS'], keep='last') # Drop duplicates just in case
with open(filename, 'wb') as f:
pickle.dump(us_combined_df, f)
def write_world_country_data(world_confirmed_df, world_deaths_df, world_population_df, filename, start_date=''):
column_names = ['Country_Region', 'Date', 'Cases', 'Deaths', 'Population']
if os.path.exists(filename):
with open(filename, 'rb') as f:
world_combined_df = pickle.load(f)
else:
world_combined_df = pd.DataFrame(columns=column_names)
if not start_date:
if len(world_combined_df):
start_date = world_combined_df['Date'].max() + timedelta(0, 0, 1)
else:
start_date = datetime(2019, 1, 1)
else:
start_date = datetime.strptime(start_date, '%Y-%m-%d')
# Use date conversions to extract correct dates from column names in COVID dataset
date_cols = [x for x in list(world_confirmed_df) if is_date(x)]
dates = [datetime.strptime(x, '%m/%d/%y') for x in date_cols]
dates = [x for x in dates if x >= start_date]
date_cols = [x for x in date_cols if datetime.strptime(x, '%m/%d/%y') >= start_date]
# Aggregate regions for countries
world_confirmed_agg_df = world_confirmed_df.groupby(['Country/Region']).agg({
d: 'sum' for d in date_cols
}).reset_index()
world_deaths_agg_df = world_deaths_df.groupby(['Country/Region']).agg({
d: 'sum' for d in date_cols
}).reset_index()
# Start populating combined dataframe to pickle
for index, row in world_confirmed_agg_df.iterrows():
country = row['Country/Region']
population = world_population_df[world_population_df['country'] == row['Country/Region']]['Population']
if len(population) != 1:
print('skipping country', country, population)
continue
population = population.to_numpy()[0]
for (date_col, date) in zip(date_cols, dates):
confirmed = row[date_col]
if confirmed == 0:
continue
if date_col in world_deaths_agg_df:
deaths = world_deaths_agg_df[
(world_deaths_agg_df['Country/Region'] == row['Country/Region'])
][date_col].to_numpy()[0]
else:
deaths = 0
values = [country, date, confirmed, deaths, population]
df_length = len(world_combined_df)
world_combined_df.loc[df_length] = values
if index % 100 == 0:
print('processed {} out of {}'.format(index, len(world_confirmed_agg_df)))
with open(filename, 'wb') as f:
pickle.dump(world_combined_df, f)
def preprocess_data(argv):
# Get covid confirmed cases and deaths for each US county and date.
us_confirmed_df = pd.read_csv(
'../COVID-19/csse_covid_19_data/csse_covid_19_time_series/time_series_covid19_confirmed_US.csv')
us_deaths_df = pd.read_csv(
'../COVID-19/csse_covid_19_data/csse_covid_19_time_series/time_series_covid19_deaths_US.csv')
# Gather covid confirmed cases and deaths for each country and date.
world_confirmed_df = pd.read_csv(
'../COVID-19/csse_covid_19_data/csse_covid_19_time_series/time_series_covid19_confirmed_global.csv')
world_deaths_df = pd.read_csv(
'../COVID-19/csse_covid_19_data/csse_covid_19_time_series/time_series_covid19_deaths_global.csv')
# US County level census data
# Modified from
# https://www2.census.gov/programs-surveys/popest/datasets/2010-2019/counties/totals/co-est2019-alldata.csv
us_county_census_df = pd.read_csv('./data/co-est2019-alldata.csv', encoding='latin-1')
# World population data (for MOST countries).
world_population_df = pd.read_csv('./data/country_profile_variables.csv')
world_population_df = world_population_df[['country', 'Population in thousands (2017)']]
world_population_df = world_population_df.rename(columns={'Population in thousands (2017)': 'Population'})
# Add some missing countries
world_population_df.loc[len(world_population_df)] = ['Taiwan*', 23780]
world_population_df['Population'] *= 1000
# Write to disk
write_us_county_data(us_confirmed_df, us_deaths_df, us_county_census_df, './data/us_combined_df.pkl',
FLAGS.start_date)
write_world_country_data(world_confirmed_df, world_deaths_df, world_population_df, './data/world_combined_df.pkl',
FLAGS.start_date)
if __name__ == "__main__":
app.run(preprocess_data)