STEP 1: GET THAT DATA

In [0]:
import pandas as pd
import numpy as np
from fbprophet import Prophet 
df_train = pd.read_csv('https://raw.githubusercontent.com/danielcaraway/data/master/covid19_0324/ca_train.csv')
df_test = pd.read_csv('https://raw.githubusercontent.com/danielcaraway/data/master/covid19_0324/ca_test.csv')
df_sub = pd.read_csv('https://raw.githubusercontent.com/danielcaraway/data/master/covid19_0324/ca_submission.csv')

STEP 2: CLEAN THAT DATA

  • For this round, we're only going to take data with confirmed cases
In [0]:
df_train = df_train[df_train['ConfirmedCases'] > 0]
In [15]:
import seaborn as sns
sns.lineplot(df_train['Date'], df_train['ConfirmedCases'])
Out[15]:
<matplotlib.axes._subplots.AxesSubplot at 0x7f191dea4048>
In [41]:
sns.regplot(df_train['Id'], np.log(df_train['ConfirmedCases']))
Out[41]:
<matplotlib.axes._subplots.AxesSubplot at 0x7f191aa5ff28>

STEP 3: MODEL THAT DATA

3A: ConfirmedCases

In [0]:
from sklearn.linear_model import LinearRegression
model_lr = LinearRegression()

X = Id; y = ConfirmedCases

In [22]:
X = np.array(df_train['Id']).reshape(-1,1)
y = df_train['ConfirmedCases']
model_lr.fit(X,y)
print('R2', model_lr.score(X,y))

R2 0.8856238923259411

X = Id; y = log(ConfirmedCases)

In [23]:
X = np.array(df_train['Id']).reshape(-1,1)
y = np.log(df_train['ConfirmedCases'])
model_lr.fit(X,y)
print('R2', model_lr.score(X,y))

R2 0.9973927213202164

3B: Fatalities

In [26]:
X = np.array(df_train['Id']).reshape(-1,1)
y = df_train['Fatalities']
model_lr.fit(X,y)
print('R2', model_lr.score(X,y))

R2 0.8583373309028113
In [25]:
X = np.array(df_train['Id']).reshape(-1,1)
y = np.log(df_train['Fatalities'])
model_lr.fit(X,y)
print('R2', model_lr.score(X,y))

R2 0.9898087710418456

3C: Confirmed Cases v Fatalities

In [27]:
X = np.array(df_train['ConfirmedCases']).reshape(-1,1)
y = df_train['Fatalities']
model_lr.fit(X,y)
print('R2', model_lr.score(X,y))

R2 0.9939565255603854
In [29]:
X = np.array(np.log(df_train['ConfirmedCases'])).reshape(-1,1)
y = np.log(df_train['Fatalities'])
model_lr.fit(X,y)
print('R2', model_lr.score(X,y))

R2 0.987548984456668

STEP 4: Pick Best Models

In [33]:
cc_model = LinearRegression() # cc = Confirmed Count
X = np.array(df_train['Id']).reshape(-1,1)
y = np.log(df_train['ConfirmedCases'])
cc_model.fit(X,y)
print('R2', cc_model.score(X,y))

R2 0.9973927213202164
In [35]:
fc_model = LinearRegression() # fc = Fatality Count
X = np.array(df_train['ConfirmedCases']).reshape(-1,1)
y = df_train['Fatalities']
fc_model.fit(X,y)
print('R2', fc_model.score(X,y))

R2 0.9939565255603854
In [36]:
df_test.head()
Out[36]:
ForecastId Province/State Country/Region Lat Long Date
0 1 California US 36.1162 -119.6816 2020-03-12
1 2 California US 36.1162 -119.6816 2020-03-13
2 3 California US 36.1162 -119.6816 2020-03-14
3 4 California US 36.1162 -119.6816 2020-03-15
4 5 California US 36.1162 -119.6816 2020-03-16
In [0]: