Skip to content

Navigation Menu

Sign in
Appearance settings

Search code, repositories, users, issues, pull requests...

Provide feedback

We read every piece of feedback, and take your input very seriously.

Saved searches

Use saved searches to filter your results more quickly
Sign up
Appearance settings

Commit 521e1c4

Browse files
Merge pull request avinashkranjan#852 from ayush-raj8/master
Corona cases Time series forecasting avinashkranjan#828 
2 parents f9557a9 + 8b4dc47 commit 521e1c4

File tree

6 files changed

+262
-0
lines changed

6 files changed

+262
-0
lines changed

‎corona cases forecasting/README.md

Lines changed: 17 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,17 @@
1+
# Corona Cases Forecasting
2+
This is a python script capable of Predicting number of covid 19 cases for next 21 days .It uses daily updated data to forecast number of fresh, death and recovered covid 19 cases for next 21 days.
3+
4+
5+
## Setup instructions
6+
Fork the repo and download it on local machine .
7+
8+
Unzip the downloaded file , go to corona cases forecasting directories.
9+
10+
Install all the libraries according to the requirements.txt file in this directory.
11+
12+
13+
## Output
14+
15+
The script Displays number of cases in near future (by default 21 days) and save the image . Sample image outputs are in the Results sub directory.
16+
17+
40.8 KB
Loading[フレーム]
33.6 KB
Loading[フレーム]
42.9 KB
Loading[フレーム]

‎corona cases forecasting/main.py

Lines changed: 240 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,240 @@
1+
# importing libraries
2+
import numpy as np
3+
import pandas as pd
4+
import matplotlib.pyplot as plt
5+
from statsmodels.tsa.arima_model import ARIMA
6+
import datetime
7+
from datetime import date
8+
import warnings
9+
warnings.filterwarnings('ignore')
10+
plt.style.use('fivethirtyeight')
11+
from pmdarima import auto_arima
12+
13+
confirmed_cases = pd.read_csv('https://raw.githubusercontent.com/CSSEGISandData/COVID-19/master/csse_covid_19_data/csse_covid_19_time_series/time_series_covid19_confirmed_global.csv')
14+
deaths_reported = pd.read_csv('https://raw.githubusercontent.com/CSSEGISandData/COVID-19/master/csse_covid_19_data/csse_covid_19_time_series/time_series_covid19_deaths_global.csv')
15+
recovered_cases = pd.read_csv('https://raw.githubusercontent.com/CSSEGISandData/COVID-19/master/csse_covid_19_data/csse_covid_19_time_series/time_series_covid19_recovered_global.csv')
16+
latest_data = pd.read_csv('https://raw.githubusercontent.com/CSSEGISandData/COVID-19/master/csse_covid_19_data/csse_covid_19_daily_reports/07-15-2020.csv')
17+
## attributes
18+
# Fetching all the columns from confirmed dataset
19+
cols = confirmed_cases.keys()
20+
# Extracting the date columns
21+
confirmed = confirmed_cases.loc[:, cols[4]:cols[-1]]
22+
deaths = deaths_reported.loc[:, cols[4]:cols[-1]]
23+
recoveries = recovered_cases.loc[:, cols[4]:cols[-1]]
24+
# Range of date
25+
dates = confirmed.keys()
26+
# Summary
27+
world_cases = []
28+
total_deaths = []
29+
mortality_rate = []
30+
recovery_rate = []
31+
total_recovered = []
32+
total_active = []
33+
# Confirmed
34+
india_cases = []
35+
# Death
36+
india_deaths = []
37+
# Recovered
38+
india_recoveries = []
39+
# Fill with the dataset
40+
for i in dates:
41+
india_cases.append(confirmed_cases[confirmed_cases['Country/Region'] == 'India'][i].sum())
42+
india_deaths.append(deaths_reported[deaths_reported['Country/Region'] == 'India'][i].sum())
43+
india_recoveries.append(recovered_cases[recovered_cases['Country/Region'] == 'India'][i].sum())
44+
45+
def daily_increase(data):
46+
d = []
47+
for i in range(len(data)):
48+
if i == 0:
49+
d.append(data[0])
50+
else:
51+
d.append(data[i]-data[i-1])
52+
return d
53+
54+
def fresh_cases_daily():
55+
#confirmed cases
56+
india_daily_increase = daily_increase(india_cases)
57+
58+
# Dates pre processing
59+
days_since_1_22 = np.array([i for i in range(len(dates))]).reshape(-1, 1)
60+
61+
days_in_future = 0
62+
future_forecast = np.array([i for i in range(len(dates)+days_in_future)]).reshape(-1, 1)
63+
64+
start = '1/22/2020'
65+
start_date = datetime.datetime.strptime(start, '%m/%d/%Y')
66+
future_forecast_dates = []
67+
for i in range(len(future_forecast)):
68+
future_forecast_dates.append((start_date + datetime.timedelta(days=i)).strftime('%m/%d/%Y'))
69+
70+
dataCovid= pd.DataFrame({ 'Dates': future_forecast_dates , 'Daily Increase':india_daily_increase })
71+
train = dataCovid[:int(0.7*(len(dataCovid)))]
72+
valid = dataCovid[int(0.7*(len(dataCovid))):]
73+
#preprocessing (since arima takes univariate series as input)
74+
train.drop('Dates',axis=1,inplace=True)
75+
valid.drop('Dates',axis=1,inplace=True)
76+
model = auto_arima(train, trace=True, error_action='ignore', suppress_warnings=True)
77+
model.fit(train)
78+
forecast = model.predict(n_periods=len(valid))
79+
forecast = pd.DataFrame(forecast,index = valid.index,columns=['Prediction'])
80+
81+
def ARIMAmodel(series, order, days = 21):
82+
# Fitting and forecast the series
83+
train = [x for x in series]
84+
model = ARIMA(train, order = order)
85+
model_fit = model.fit(disp=0)
86+
forecast, err, ci = model_fit.forecast(steps = days, alpha = 0.05)
87+
start_day = date.today() + datetime.timedelta(days = 1)
88+
predictions_df = pd.DataFrame({'Forecast':forecast.round()}, index=pd.date_range(start = start_day, periods=days, freq='D'))
89+
return predictions_df, ci
90+
91+
new_positives = dataCovid['Daily Increase'].values
92+
order = {
93+
'new_positives': (2, 1, 5),
94+
}
95+
new_positives_today=new_positives[-1]
96+
# Forecasting with ARIMA models
97+
new_positives_pred, new_positives_ci = ARIMAmodel(new_positives, order['new_positives'])
98+
casesY=[]
99+
datesX=[]
100+
list1 = new_positives_pred.iloc[: ,0]
101+
for i in range(0,21):
102+
casesY.append(list1[i])
103+
datesX.append((date.today()+ datetime.timedelta(days=i)).strftime('%m/%d/%Y'))
104+
105+
# Plot Results for forecasted dates only (detailed)
106+
plt.plot(datesX,casesY,color='red')
107+
plt.title('New active Cases Forecast')
108+
plt.xticks(rotation=90)
109+
# plt.figure(figsize=(22,22))
110+
plt.savefig("./corona cases forecasting/Results/plot1.png",bbox_inches='tight')
111+
plt.autoscale()
112+
plt.show()
113+
114+
def death_cases_daily():
115+
#confirmed cases
116+
india_daily_increase = daily_increase(india_deaths)
117+
118+
# Dates pre processing
119+
days_since_1_22 = np.array([i for i in range(len(dates))]).reshape(-1, 1)
120+
121+
days_in_future = 0
122+
future_forecast = np.array([i for i in range(len(dates)+days_in_future)]).reshape(-1, 1)
123+
124+
start = '1/22/2020'
125+
start_date = datetime.datetime.strptime(start, '%m/%d/%Y')
126+
future_forecast_dates = []
127+
for i in range(len(future_forecast)):
128+
future_forecast_dates.append((start_date + datetime.timedelta(days=i)).strftime('%m/%d/%Y'))
129+
130+
dataCovid= pd.DataFrame({ 'Dates': future_forecast_dates , 'Daily Increase':india_daily_increase })
131+
train = dataCovid[:int(0.7*(len(dataCovid)))]
132+
valid = dataCovid[int(0.7*(len(dataCovid))):]
133+
#preprocessing (since arima takes univariate series as input)
134+
train.drop('Dates',axis=1,inplace=True)
135+
valid.drop('Dates',axis=1,inplace=True)
136+
model = auto_arima(train, trace=True, error_action='ignore', suppress_warnings=True)
137+
model.fit(train)
138+
forecast = model.predict(n_periods=len(valid))
139+
forecast = pd.DataFrame(forecast,index = valid.index,columns=['Prediction'])
140+
141+
def ARIMAmodel(series, order, days = 21):
142+
# Fitting and forecast the series
143+
train = [x for x in series]
144+
model = ARIMA(train, order = order)
145+
model_fit = model.fit(disp=0)
146+
forecast, err, ci = model_fit.forecast(steps = days, alpha = 0.05)
147+
start_day = date.today() + datetime.timedelta(days = 1)
148+
predictions_df = pd.DataFrame({'Forecast':forecast.round()}, index=pd.date_range(start = start_day, periods=days, freq='D'))
149+
return predictions_df, ci
150+
151+
new_deaths = dataCovid['Daily Increase'].values
152+
order = {
153+
'new_deaths': (0, 1, 1),
154+
}
155+
new_deaths_today=new_deaths[-1]
156+
# Forecasting with ARIMA models
157+
new_deaths_pred, new_deaths_ci = ARIMAmodel(new_deaths, order['new_deaths'])
158+
casesY=[]
159+
datesX=[]
160+
list1 = new_deaths_pred.iloc[: ,0]
161+
for i in range(0,21):
162+
casesY.append(list1[i])
163+
datesX.append((date.today()+ datetime.timedelta(days=i)).strftime('%m/%d/%Y'))
164+
165+
# Plot Results for forecasted dates only (detailed)
166+
plt.plot(datesX,casesY,color='red')
167+
plt.title('New death Cases Forecast')
168+
plt.xticks(rotation=90)
169+
# plt.figure(figsize=(22,22))
170+
plt.savefig("./corona cases forecasting/Results/plot2.png",bbox_inches='tight')
171+
plt.autoscale()
172+
plt.show()
173+
174+
def recovered_cases_daily():
175+
#confirmed cases
176+
india_daily_increase = daily_increase(india_recoveries)
177+
# Dates pre processing
178+
days_since_1_22 = np.array([i for i in range(len(dates))]).reshape(-1, 1)
179+
days_in_future = 0
180+
future_forecast = np.array([i for i in range(len(dates)+days_in_future)]).reshape(-1, 1)
181+
start = '1/22/2020'
182+
start_date = datetime.datetime.strptime(start, '%m/%d/%Y')
183+
future_forecast_dates = []
184+
for i in range(len(future_forecast)):
185+
future_forecast_dates.append((start_date + datetime.timedelta(days=i)).strftime('%m/%d/%Y'))
186+
187+
dataCovid= pd.DataFrame({ 'Dates': future_forecast_dates , 'Daily recoveries':india_daily_increase })
188+
train = dataCovid[:int(0.7*(len(dataCovid)))]
189+
valid = dataCovid[int(0.7*(len(dataCovid))):]
190+
#preprocessing (since arima takes univariate series as input)
191+
train.drop('Dates',axis=1,inplace=True)
192+
valid.drop('Dates',axis=1,inplace=True)
193+
model = auto_arima(train, trace=True, error_action='ignore', suppress_warnings=True)
194+
model.fit(train)
195+
forecast = model.predict(n_periods=len(valid))
196+
forecast = pd.DataFrame(forecast,index = valid.index,columns=['Prediction'])
197+
198+
def ARIMAmodel(series, order, days = 21):
199+
# Fitting and forecast the series
200+
train = [x for x in series]
201+
model = ARIMA(train, order = order)
202+
model_fit = model.fit(disp=0)
203+
forecast, err, ci = model_fit.forecast(steps = days, alpha = 0.05)
204+
start_day = date.today() + datetime.timedelta(days = 1)
205+
predictions_df = pd.DataFrame({'Forecast':forecast.round()}, index=pd.date_range(start = start_day, periods=days, freq='D'))
206+
return predictions_df, ci
207+
208+
new_recoveries = dataCovid['Daily recoveries'].values
209+
order = {
210+
'new_recoveries': (1, 1, 2),
211+
}
212+
new_recoveries_today=new_recoveries[-1]
213+
# Forecasting with ARIMA models
214+
new_recoveries_pred, new_recoveries_ci = ARIMAmodel(new_recoveries, order['new_recoveries'])
215+
casesY=[]
216+
datesX=[]
217+
list1 = new_recoveries_pred.iloc[: ,0]
218+
for i in range(0,21):
219+
casesY.append(list1[i])
220+
datesX.append((date.today()+ datetime.timedelta(days=i)).strftime('%m/%d/%Y'))
221+
222+
# Plot Results for forecasted dates only (detailed)
223+
plt.plot(datesX,casesY,color='red')
224+
plt.title('New recovered Cases Forecast')
225+
plt.xticks(rotation=90)
226+
# plt.figure(figsize=(22,22))
227+
plt.savefig("./corona cases forecasting/Results/plot3.png",bbox_inches='tight')
228+
plt.autoscale()
229+
plt.show()
230+
231+
# Taking user input choice for type of prediction method to be intitiated
232+
choice=input("F for fresh cases,D for death cases,R for recovered cases prediction : ")
233+
if choice=='F':
234+
fresh_cases_daily()
235+
elif choice=='D':
236+
death_cases_daily()
237+
elif choice=='R':
238+
recovered_cases_daily()
239+
else:
240+
print("Enter a valid choice")
Lines changed: 5 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,5 @@
1+
numpy==1.19.2
2+
pandas==1.1.3
3+
matplotlib==3.3.2
4+
statsmodels==0.12.2
5+
pmdarima==1.8.0

0 commit comments

Comments
(0)

AltStyle によって変換されたページ (->オリジナル) /