
# import libraries
import datetime
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns
import tensorflow as tf
import pmdarima as pm
import itertools
import statsmodels.api as sm
import random

from statsmodels.tsa.arima_model import ARIMA

from sklearn.preprocessing import StandardScaler, MinMaxScaler
from sklearn.metrics import r2_score, mean_squared_error
from sklearn.model_selection import train_test_split, TimeSeriesSplit

from tensorflow.keras.models import Sequential, load_model
from tensorflow.keras.layers import Dense, Dropout, GRU, LSTM
from tensorflow.keras.optimizers import Adam
from tensorflow.keras.regularizers import l2
from tensorflow.keras.preprocessing.sequence import TimeseriesGenerator
from tensorflow.keras.callbacks import EarlyStopping
from tensorflow.random import set_seed

import warnings
warnings.filterwarnings('ignore')

gpus = tf.config.experimental.list_physical_devices('GPU')
tf.config.experimental.set_memory_growth(gpus[0], True)

%matplotlib inline

random.seed(42)
set_seed(42)


# import datasets
cases_df = pd.read_csv('../clean_data/Cases_Normalized_per_100k.csv')
cases_df.head()


cases_df.describe()


# convert date to datetime
cases_df['date'] = pd.to_datetime(cases_df['date'])


# drop no longer needed columns
cases_df.drop(['totalcountconfirmed', 'totalcountdeaths', 'newcountconfirmed',
              'newcountdeaths', 'NEVER', 'RARELY', 'SOMETIMES', 'FREQUENTLY', 
              'ALWAYS', 'deathper100k', 'newdeath100k', 'confirmedper100k', 'population',
               'newlyconfirmedper100k' ,'7dayrollingavg_newlyconfirmed_rawnumber', '7dayrollingavg_newdeath_rawnumber'], 
             axis = 1, inplace = True)
cases_df.head()


# bring in hospital data
hospitals_df = pd.read_csv('../clean_data/hospitals_by_county.csv')
hospitals_df.head()


# convert to datetime
hospitals_df['todays_date'] = pd.to_datetime(hospitals_df['todays_date'])


# drop unncessary columns
hospitals_df.drop(['hospitalized_covid_confirmed_patients', 'hospitalized_suspected_covid_patients', 
                   'icu_covid_confirmed_patients', 'icu_suspected_covid_patients'],
                 axis = 1, inplace = True)
hospitals_df.head()


cases_hosp_df = cases_df.merge(hospitals_df, how = 'inner', left_on = ['county','date'], right_on = ['county', 'todays_date'])
cases_hosp_df.head()


# do a group by so we can create predictions per county
county_df = cases_hosp_df.groupby(['county', 'date']).mean()


county_df.index.get_level_values('county').unique()[0]

'Alameda'


# start by analyzing LA County
county_df.loc['Los Angeles']


county_df.to_csv('../clean_data/counties_models.csv')


la_df = county_df.loc['Los Angeles']
features = ['hospitalized_covid_patients']
X = la_df[features]
y = la_df[['7dayrollingavg_newlyconfirmed']].values


# Train/Test Split
X_train, X_test, y_train, y_test = train_test_split(X, y, shuffle = False, test_size = 0.15)


# scale our data
mms = MinMaxScaler()
X_train_sc = mms.fit_transform(X_train)
X_test_sc = mms.transform(X_test)


# Create training sequences
train_sequences = TimeseriesGenerator(X_train_sc, y_train, length = 1,  batch_size = 32)


# Create test sequences
batch_x, batch_y = train_sequences[0]


test_sequences = TimeseriesGenerator(X_test_sc, y_test, length = 1,  batch_size = 32)


input_shape = train_sequences[0][0][0].shape


# Design RNN
model = Sequential()
model.add(LSTM(128, input_shape = (input_shape), return_sequences = True))
model.add(Dropout(0.2))
model.add(LSTM(64, return_sequences = True))
model.add(Dropout(0.2))
model.add(Dense(32, activation = 'swish', kernel_regularizer = l2(0.001)))
model.add(Dropout(0.2))
model.add(Dense(16, activation = 'swish', kernel_regularizer = l2(0.001)))
model.add(Dense(1))


model.compile(optimizer=Adam(lr = 0.005), loss = 'mse')
early_stop = EarlyStopping(monitor = 'val_loss', patience = 20, verbose = 1, mode = 'auto')


history = model.fit(train_sequences, validation_data = (test_sequences), epochs = 200, verbose = 0,callbacks = [early_stop] )

Epoch 00125: early stopping


print(history.history['loss'][-1]** 0.5)
print(history.history['val_loss'][-1] ** 0.5)

8.408704661653545
16.79715297735099


preds_train = model.predict(train_sequences)


preds_train.shape

(255, 1, 1)


train_preds = preds_train[:,0]


train_preds = train_preds.reshape(255)


preds_test = model.predict(test_sequences)


preds_test.shape

(45, 1, 1)


test_preds = preds_test[:, 0]
test_preds = test_preds.reshape(45)


test_preds.shape

(45,)


plt.figure(figsize=(15,15))
sns.lineplot(x = range(0, len(history.history['loss'])), y = [i ** 0.5 for i in history.history['loss']], label = 'Train Loss')
sns.lineplot(x = range(0, len(history.history['loss'])), y = [i ** 0.5 for i in history.history['val_loss']], label = 'Test Loss')
plt.title('Train vs Test Loss for Los Angeles County', fontsize = 20, fontweight = 'bold')
plt.xlabel('Epochs')
plt.ylabel('Root Mean Squared Error')
plt.legend()
plt.savefig('../images/train_test_loss_la_county.png');


# generate predictions
future_pred_count = 1
future = []
batch = X_test_sc[-future_pred_count:].reshape((1, future_pred_count, 1))
model.predict(batch)[0].shape

for i in range(22):
    pred = model.predict(batch)[0]
    future.append(pred)
    batch = np.append(batch[:, 0:, :], [[pred[i]]] , axis = 1)


dates = [la_df.index[-1] + pd.tseries.offsets.DateOffset(days = x) for x in range(0, 23)]
future_date = pd.DataFrame(index = dates[1:], columns = X.columns)
future_date.tail()


df_predict = pd.DataFrame(future[21], index = future_date[-22:].index, columns = ['Prediction'])
df_predict.head()


df_test = pd.concat([la_df, df_predict], axis = 1)
df_test


plt.figure(figsize=(15, 5))
plt.plot(df_test.index, df_test['7dayrollingavg_newlyconfirmed'])
plt.plot(df_test.index, df_test['Prediction'], color = 'r')
plt.plot(X_train.index[:-1], train_preds, color = 'g')
plt.plot(X_test.index[:-1], test_preds, color = 'g')
plt.legend(['7 Day Rolling Average Newly Confirmed', 'Future Prediction', 'Current Time Prediction'], loc = 'best', fontsize = 'xx-large')
plt.xticks(fontsize = 12)
plt.yticks(fontsize = 16)
plt.xlabel('Date')
plt.ylabel('7 Day Rolloing Average Cases per 100k People')
plt.title('Prediction of 7 Day Rolling Average Cases per 100k People in LA County', fontsize = 20, fontweight = 'bold')
plt.savefig('../images/rolling_7_day_prediction.png');


la_df = county_df.loc['Los Angeles'].tail(110)
features = ['hospitalized_covid_patients', '7dayrollingavg_newdeath', 'all_hospital_beds', 'icu_available_beds']
X = la_df[features]
y = la_df[['7dayrollingavg_newlyconfirmed']]


X_train = X.iloc[:92]
X_test = X.iloc[92:]
y_train = y.iloc[:92]
y_test = y.iloc[92:]
X_train.tail()


smodel = pm.auto_arima(y_train, start_p=1, start_q=1,
                         test='adf',
                         max_p=3, max_q=3, m=12,
                         start_P=0, seasonal=False,
                         d=None, D=None, trace=True,
                         error_action='ignore',  
                         suppress_warnings=True, 
                         stepwise=True)

smodel.summary()

Performing stepwise search to minimize aic
 ARIMA(1,2,1)(0,0,0)[0] intercept   : AIC=575.359, Time=0.11 sec
 ARIMA(0,2,0)(0,0,0)[0] intercept   : AIC=649.159, Time=0.01 sec
 ARIMA(1,2,0)(0,0,0)[0] intercept   : AIC=611.739, Time=0.04 sec
 ARIMA(0,2,1)(0,0,0)[0] intercept   : AIC=inf, Time=0.10 sec
 ARIMA(0,2,0)(0,0,0)[0]             : AIC=647.160, Time=0.01 sec
 ARIMA(2,2,1)(0,0,0)[0] intercept   : AIC=576.989, Time=0.07 sec
 ARIMA(1,2,2)(0,0,0)[0] intercept   : AIC=576.229, Time=0.09 sec
 ARIMA(0,2,2)(0,0,0)[0] intercept   : AIC=574.601, Time=0.12 sec
 ARIMA(0,2,3)(0,0,0)[0] intercept   : AIC=576.291, Time=0.08 sec
 ARIMA(1,2,3)(0,0,0)[0] intercept   : AIC=577.940, Time=0.14 sec
 ARIMA(0,2,2)(0,0,0)[0]             : AIC=572.608, Time=0.05 sec
 ARIMA(0,2,1)(0,0,0)[0]             : AIC=575.638, Time=0.02 sec
 ARIMA(1,2,2)(0,0,0)[0]             : AIC=574.247, Time=0.06 sec
 ARIMA(0,2,3)(0,0,0)[0]             : AIC=574.308, Time=0.05 sec
 ARIMA(1,2,1)(0,0,0)[0]             : AIC=573.362, Time=0.04 sec
 ARIMA(1,2,3)(0,0,0)[0]             : AIC=575.941, Time=0.10 sec

Best model:  ARIMA(0,2,2)(0,0,0)[0]          
Total fit time: 1.117 seconds


plt_smodel = smodel.plot_diagnostics(figsize=(16, 8))
plt_smodel.show()
plt_smodel.savefig('../images/arima_model_diags.png')


warnings.filterwarnings("ignore") 

model = ARIMA(y_train, order=smodel.order)  
fitted = model.fit(disp=-1)  
print(fitted.summary())
# Forecast
fc, se, conf = fitted.forecast(18, alpha=0.05)  # 95% conf

# Make as pandas series
fc_series = pd.Series(fc, index=y_test.index)
lower_series = pd.Series(conf[:, 0], index=y_test.index)
upper_series = pd.Series(conf[:, 1], index=y_test.index)

# Plot
plt.figure(figsize=(14,5))
plt.plot(y_train, label='training')
plt.plot(y_test, label='actual')
plt.plot(fc_series, label='forecast')
plt.fill_between(lower_series.index, lower_series, upper_series, 
                 color='k', alpha=.15)
plt.title('Forecast vs Actuals')
plt.legend()
plt.savefig('../images/arima_la_county.png');

                                    ARIMA Model Results                                     
============================================================================================
Dep. Variable:     D2.7dayrollingavg_newlyconfirmed   No. Observations:                   90
Model:                               ARIMA(0, 2, 2)   Log Likelihood                -283.300
Method:                                     css-mle   S.D. of innovations              5.567
Date:                              Fri, 29 Jan 2021   AIC                            574.601
Time:                                      17:42:59   BIC                            584.600
Sample:                                  10-09-2020   HQIC                           578.633
                                       - 01-06-2021                                         
==========================================================================================================
                                             coef    std err          z      P>|z|      [0.025      0.975]
----------------------------------------------------------------------------------------------------------
const                                     -0.0050      0.059     -0.085      0.932      -0.121       0.111
ma.L1.D2.7dayrollingavg_newlyconfirmed    -1.1886      0.110    -10.791      0.000      -1.405      -0.973
ma.L2.D2.7dayrollingavg_newlyconfirmed     0.2793      0.118      2.377      0.017       0.049       0.510
                                    Roots                                    
=============================================================================
                  Real          Imaginary           Modulus         Frequency
-----------------------------------------------------------------------------
MA.1            1.1545           +0.0000j            1.1545            0.0000
MA.2            3.1012           +0.0000j            3.1012            0.0000
-----------------------------------------------------------------------------


print('MSE of our forecast is:', mean_squared_error(y_test, fc_series, squared=True))
print('RMSE of our forecast is:', mean_squared_error(y_test, fc_series, squared=False))

MSE of our forecast is: 524.1637136663361
RMSE of our forecast is: 22.89462193761531


smodel = pm.auto_arima(y, start_p=1, start_q=1,
                         test='adf',
                         max_p=3, max_q=3, m=12,
                         start_P=0, seasonal=False,
                         d=None, D=None, trace=True,
                         error_action='ignore',  
                         suppress_warnings=True, 
                         stepwise=True)

smodel.summary()

Performing stepwise search to minimize aic
 ARIMA(1,2,1)(0,0,0)[0] intercept   : AIC=684.987, Time=0.05 sec
 ARIMA(0,2,0)(0,0,0)[0] intercept   : AIC=761.359, Time=0.01 sec
 ARIMA(1,2,0)(0,0,0)[0] intercept   : AIC=718.901, Time=0.03 sec
 ARIMA(0,2,1)(0,0,0)[0] intercept   : AIC=686.186, Time=0.04 sec
 ARIMA(0,2,0)(0,0,0)[0]             : AIC=759.361, Time=0.01 sec
 ARIMA(2,2,1)(0,0,0)[0] intercept   : AIC=686.897, Time=0.07 sec
 ARIMA(1,2,2)(0,0,0)[0] intercept   : AIC=686.522, Time=0.09 sec
 ARIMA(0,2,2)(0,0,0)[0] intercept   : AIC=684.739, Time=0.06 sec
 ARIMA(0,2,3)(0,0,0)[0] intercept   : AIC=686.600, Time=0.08 sec
 ARIMA(1,2,3)(0,0,0)[0] intercept   : AIC=687.512, Time=0.16 sec
 ARIMA(0,2,2)(0,0,0)[0]             : AIC=683.386, Time=0.03 sec
 ARIMA(0,2,1)(0,0,0)[0]             : AIC=684.809, Time=0.02 sec
 ARIMA(1,2,2)(0,0,0)[0]             : AIC=685.147, Time=0.04 sec
 ARIMA(0,2,3)(0,0,0)[0]             : AIC=685.232, Time=0.05 sec
 ARIMA(1,2,1)(0,0,0)[0]             : AIC=683.639, Time=0.03 sec
 ARIMA(1,2,3)(0,0,0)[0]             : AIC=686.161, Time=0.12 sec

Best model:  ARIMA(0,2,2)(0,0,0)[0]          
Total fit time: 0.921 seconds


# Forecast
n_periods = 3
fitted, confint = smodel.predict(n_periods=n_periods, return_conf_int=True)
index_of_fc = pd.date_range(y.index[-1], periods=n_periods, freq='W')

# make series for plotting purpose
fitted_series = pd.Series(fitted, index=index_of_fc)
lower_series = pd.Series(confint[:, 0], index=index_of_fc)
upper_series = pd.Series(confint[:, 1], index=index_of_fc)

# Plot

fig = plt.figure(figsize=(14, 5))

plt.plot(y, label="Actual")
plt.plot(fitted_series, color='orange', label='Forecast')
plt.fill_between(lower_series.index,
                 lower_series,
                 upper_series,
                 color='k', alpha=.1)

plt.title('Final Forecast')
plt.legend()
plt.show()


hotspot_counties = ['Tehama', 'Riverside', 'Colusa', 'Merced', 'Santa Barbara', 'Inyo', 'Tulare']

features = ['hospitalized_covid_patients', '7dayrollingavg_newdeath', 'all_hospital_beds', 'icu_available_beds']


## Tehama County

Tehama = county_df.loc['Tehama'].tail(110)
Riverside = county_df.loc['Riverside'].tail(110)
Colusa = county_df.loc['Colusa'].tail(110)
Merced = county_df.loc['Merced'].tail(110)
Santa_Barbara = county_df.loc['Santa Barbara'].tail(110)
Inyo = county_df.loc['Inyo'].tail(110)
Tulare = county_df.loc['Tulare'].tail(110)


# build a function to incorporate everything 
warnings.filterwarnings("ignore") 
def arima_model (countydf):
    X = countydf[features]
    y = countydf[['7dayrollingavg_newlyconfirmed']]   
    X_train = X.iloc[:92]
    X_test = X.iloc[92:]
    y_train = y.iloc[:92]
    y_test = y.iloc[92:]
    
    smodel = pm.auto_arima(y_train, start_p=1, start_q=1,
                         test='adf',
                         max_p=3, max_q=3, m=12,
                         start_P=0, seasonal=False,
                         d=None, D=None, trace=True,
                         error_action='ignore',  
                         suppress_warnings=True, 
                         stepwise=True)

    print(smodel.summary())

    model = ARIMA(y_train, order=smodel.order)  
    fitted = model.fit(disp=-1)  
    print(fitted.summary())
    # Forecast
    fc, se, conf = fitted.forecast(18, alpha=0.05)  # 95% conf

    # Make as pandas series
    fc_series = pd.Series(fc, index=y_test.index)
    lower_series = pd.Series(conf[:, 0], index=y_test.index)
    upper_series = pd.Series(conf[:, 1], index=y_test.index)

    # Plot
    plt.figure(figsize=(14,5))
    plt.plot(y_train, label='training')
    plt.plot(y_test, label='actual')
    plt.plot(fc_series, label='forecast')
    plt.fill_between(lower_series.index, lower_series, upper_series, 
                     color='k', alpha=.15)
    plt.title('Forecast vs Actuals')
    plt.legend()
    plt.show()



    print('MSE of our forecast is:', mean_squared_error(y_test, fc_series, squared=True))
    print('RMSE of our forecast is:', mean_squared_error(y_test, fc_series, squared=False))


    
    # use the whole dataset to make actual predictions
    
    smodel = pm.auto_arima(y, start_p=1, start_q=1,
                             test='adf',
                             max_p=3, max_q=3, m=12,
                             start_P=0, seasonal=False,
                             d=None, D=None, trace=True,
                             error_action='ignore',  
                             suppress_warnings=True, 
                             stepwise=True)

    print(smodel.summary())


    # Forecast
    n_periods = 3
    fitted, confint = smodel.predict(n_periods=n_periods, return_conf_int=True)
    index_of_fc = pd.date_range(y.index[-1], periods=n_periods, freq='W')

    # make series for plotting purpose
    fitted_series = pd.Series(fitted, index=index_of_fc)
    lower_series = pd.Series(confint[:, 0], index=index_of_fc)
    upper_series = pd.Series(confint[:, 1], index=index_of_fc)

    # Plot

    fig = plt.figure(figsize=(14, 5))

    plt.plot(y, label="Actual")
    plt.plot(fitted_series, color='orange', label='Forecast')
    plt.fill_between(lower_series.index,
                     lower_series,
                     upper_series,
                     color='k', alpha=.1)

    plt.title('Final Forecast')
    plt.legend()
    plt.show()
    print(fitted_series)
    return


arima_model(Tehama)

Performing stepwise search to minimize aic
 ARIMA(1,2,1)(0,0,0)[0] intercept   : AIC=inf, Time=0.14 sec
 ARIMA(0,2,0)(0,0,0)[0] intercept   : AIC=596.915, Time=0.01 sec
 ARIMA(1,2,0)(0,0,0)[0] intercept   : AIC=567.628, Time=0.03 sec
 ARIMA(0,2,1)(0,0,0)[0] intercept   : AIC=inf, Time=0.09 sec
 ARIMA(0,2,0)(0,0,0)[0]             : AIC=594.944, Time=0.01 sec
 ARIMA(2,2,0)(0,0,0)[0] intercept   : AIC=554.762, Time=0.04 sec
 ARIMA(3,2,0)(0,0,0)[0] intercept   : AIC=547.391, Time=0.06 sec
 ARIMA(3,2,1)(0,0,0)[0] intercept   : AIC=543.793, Time=0.09 sec
 ARIMA(2,2,1)(0,0,0)[0] intercept   : AIC=inf, Time=0.18 sec
 ARIMA(3,2,2)(0,0,0)[0] intercept   : AIC=inf, Time=0.29 sec
 ARIMA(2,2,2)(0,0,0)[0] intercept   : AIC=inf, Time=0.17 sec
 ARIMA(3,2,1)(0,0,0)[0]             : AIC=542.016, Time=0.06 sec
 ARIMA(2,2,1)(0,0,0)[0]             : AIC=inf, Time=0.13 sec
 ARIMA(3,2,0)(0,0,0)[0]             : AIC=545.566, Time=0.03 sec
 ARIMA(3,2,2)(0,0,0)[0]             : AIC=inf, Time=0.25 sec
 ARIMA(2,2,0)(0,0,0)[0]             : AIC=552.906, Time=0.02 sec
 ARIMA(2,2,2)(0,0,0)[0]             : AIC=inf, Time=0.12 sec

Best model:  ARIMA(3,2,1)(0,0,0)[0]          
Total fit time: 1.724 seconds
                               SARIMAX Results                                
==============================================================================
Dep. Variable:                      y   No. Observations:                   92
Model:               SARIMAX(3, 2, 1)   Log Likelihood                -266.008
Date:                Fri, 29 Jan 2021   AIC                            542.016
Time:                        17:43:02   BIC                            554.515
Sample:                             0   HQIC                           547.056
                                 - 92                                         
Covariance Type:                  opg                                         
==============================================================================
                 coef    std err          z      P>|z|      [0.025      0.975]
------------------------------------------------------------------------------
ar.L1         -0.4776      0.325     -1.472      0.141      -1.114       0.159
ar.L2         -0.3546      0.216     -1.638      0.101      -0.779       0.070
ar.L3         -0.1883      0.184     -1.024      0.306      -0.549       0.172
ma.L1         -0.5083      0.262     -1.937      0.053      -1.023       0.006
sigma2        21.3068      3.318      6.422      0.000      14.804      27.809
===================================================================================
Ljung-Box (L1) (Q):                   0.06   Jarque-Bera (JB):                 2.71
Prob(Q):                              0.80   Prob(JB):                         0.26
Heteroskedasticity (H):               7.35   Skew:                            -0.42
Prob(H) (two-sided):                  0.00   Kurtosis:                         3.17
===================================================================================

Warnings:
[1] Covariance matrix calculated using the outer product of gradients (complex-step).
                                    ARIMA Model Results                                     
============================================================================================
Dep. Variable:     D2.7dayrollingavg_newlyconfirmed   No. Observations:                   90
Model:                               ARIMA(3, 2, 1)   Log Likelihood                -265.896
Method:                                     css-mle   S.D. of innovations              4.610
Date:                              Fri, 29 Jan 2021   AIC                            543.793
Time:                                      17:43:02   BIC                            558.792
Sample:                                  10-09-2020   HQIC                           549.841
                                       - 01-06-2021                                         
==========================================================================================================
                                             coef    std err          z      P>|z|      [0.025      0.975]
----------------------------------------------------------------------------------------------------------
const                                      0.0570      0.121      0.471      0.638      -0.180       0.294
ar.L1.D2.7dayrollingavg_newlyconfirmed    -0.4783      0.204     -2.342      0.019      -0.879      -0.078
ar.L2.D2.7dayrollingavg_newlyconfirmed    -0.3548      0.185     -1.914      0.056      -0.718       0.009
ar.L3.D2.7dayrollingavg_newlyconfirmed    -0.1883      0.143     -1.313      0.189      -0.469       0.093
ma.L1.D2.7dayrollingavg_newlyconfirmed    -0.5090      0.194     -2.624      0.009      -0.889      -0.129
                                    Roots                                    
=============================================================================
                  Real          Imaginary           Modulus         Frequency
-----------------------------------------------------------------------------
AR.1            0.0410           -1.6429j            1.6434           -0.2460
AR.2            0.0410           +1.6429j            1.6434            0.2460
AR.3           -1.9658           -0.0000j            1.9658           -0.5000
MA.1            1.9647           +0.0000j            1.9647            0.0000
-----------------------------------------------------------------------------

MSE of our forecast is: 3665.4470207701274
RMSE of our forecast is: 60.54293534980054
Performing stepwise search to minimize aic
 ARIMA(1,1,1)(0,0,0)[0] intercept   : AIC=653.938, Time=0.05 sec
 ARIMA(0,1,0)(0,0,0)[0] intercept   : AIC=654.595, Time=0.01 sec
 ARIMA(1,1,0)(0,0,0)[0] intercept   : AIC=652.013, Time=0.03 sec
 ARIMA(0,1,1)(0,0,0)[0] intercept   : AIC=652.508, Time=0.04 sec
 ARIMA(0,1,0)(0,0,0)[0]             : AIC=653.046, Time=0.01 sec
 ARIMA(2,1,0)(0,0,0)[0] intercept   : AIC=653.909, Time=0.04 sec
 ARIMA(2,1,1)(0,0,0)[0] intercept   : AIC=655.115, Time=0.10 sec
 ARIMA(1,1,0)(0,0,0)[0]             : AIC=650.691, Time=0.01 sec
 ARIMA(2,1,0)(0,0,0)[0]             : AIC=652.523, Time=0.02 sec
 ARIMA(1,1,1)(0,0,0)[0]             : AIC=652.574, Time=0.02 sec
 ARIMA(0,1,1)(0,0,0)[0]             : AIC=651.216, Time=0.02 sec
 ARIMA(2,1,1)(0,0,0)[0]             : AIC=653.454, Time=0.06 sec

Best model:  ARIMA(1,1,0)(0,0,0)[0]          
Total fit time: 0.435 seconds
                               SARIMAX Results                                
==============================================================================
Dep. Variable:                      y   No. Observations:                  110
Model:               SARIMAX(1, 1, 0)   Log Likelihood                -323.346
Date:                Fri, 29 Jan 2021   AIC                            650.691
Time:                        17:43:03   BIC                            656.074
Sample:                             0   HQIC                           652.874
                                - 110                                         
Covariance Type:                  opg                                         
==============================================================================
                 coef    std err          z      P>|z|      [0.025      0.975]
------------------------------------------------------------------------------
ar.L1         -0.1970      0.093     -2.118      0.034      -0.379      -0.015
sigma2        22.0807      2.690      8.210      0.000      16.809      27.352
===================================================================================
Ljung-Box (L1) (Q):                   0.00   Jarque-Bera (JB):                 1.64
Prob(Q):                              0.99   Prob(JB):                         0.44
Heteroskedasticity (H):               6.74   Skew:                            -0.14
Prob(H) (two-sided):                  0.00   Kurtosis:                         3.54
===================================================================================

Warnings:
[1] Covariance matrix calculated using the outer product of gradients (complex-step).

2021-01-24    49.300240
2021-01-31    49.317279
2021-02-07    49.313922
Freq: W-SUN, dtype: float64


## Riverside County

arima_model(Riverside)

Performing stepwise search to minimize aic
 ARIMA(1,1,1)(0,0,0)[0] intercept   : AIC=563.188, Time=0.08 sec
 ARIMA(0,1,0)(0,0,0)[0] intercept   : AIC=581.413, Time=0.01 sec
 ARIMA(1,1,0)(0,0,0)[0] intercept   : AIC=561.905, Time=0.03 sec
 ARIMA(0,1,1)(0,0,0)[0] intercept   : AIC=564.721, Time=0.03 sec
 ARIMA(0,1,0)(0,0,0)[0]             : AIC=584.545, Time=0.01 sec
 ARIMA(2,1,0)(0,0,0)[0] intercept   : AIC=562.932, Time=0.05 sec
 ARIMA(2,1,1)(0,0,0)[0] intercept   : AIC=564.921, Time=0.07 sec
 ARIMA(1,1,0)(0,0,0)[0]             : AIC=561.854, Time=0.02 sec
 ARIMA(2,1,0)(0,0,0)[0]             : AIC=563.251, Time=0.03 sec
 ARIMA(1,1,1)(0,0,0)[0]             : AIC=563.352, Time=0.03 sec
 ARIMA(0,1,1)(0,0,0)[0]             : AIC=566.089, Time=0.02 sec
 ARIMA(2,1,1)(0,0,0)[0]             : AIC=565.251, Time=0.03 sec

Best model:  ARIMA(1,1,0)(0,0,0)[0]          
Total fit time: 0.426 seconds
                               SARIMAX Results                                
==============================================================================
Dep. Variable:                      y   No. Observations:                   92
Model:               SARIMAX(1, 1, 0)   Log Likelihood                -278.927
Date:                Fri, 29 Jan 2021   AIC                            561.854
Time:                        17:43:04   BIC                            566.875
Sample:                             0   HQIC                           563.880
                                 - 92                                         
Covariance Type:                  opg                                         
==============================================================================
                 coef    std err          z      P>|z|      [0.025      0.975]
------------------------------------------------------------------------------
ar.L1          0.4864      0.069      7.084      0.000       0.352       0.621
sigma2        26.8271      2.167     12.381      0.000      22.580      31.074
===================================================================================
Ljung-Box (L1) (Q):                   0.07   Jarque-Bera (JB):               225.20
Prob(Q):                              0.80   Prob(JB):                         0.00
Heteroskedasticity (H):              56.29   Skew:                            -0.61
Prob(H) (two-sided):                  0.00   Kurtosis:                        10.61
===================================================================================

Warnings:
[1] Covariance matrix calculated using the outer product of gradients (complex-step).
                                    ARIMA Model Results                                    
===========================================================================================
Dep. Variable:     D.7dayrollingavg_newlyconfirmed   No. Observations:                   91
Model:                              ARIMA(1, 1, 0)   Log Likelihood                -277.953
Method:                                    css-mle   S.D. of innovations              5.125
Date:                             Fri, 29 Jan 2021   AIC                            561.905
Time:                                     17:43:04   BIC                            569.438
Sample:                                 10-08-2020   HQIC                           564.944
                                      - 01-06-2021                                         
=========================================================================================================
                                            coef    std err          z      P>|z|      [0.025      0.975]
---------------------------------------------------------------------------------------------------------
const                                     1.4125      0.981      1.440      0.150      -0.510       3.335
ar.L1.D.7dayrollingavg_newlyconfirmed     0.4571      0.092      4.943      0.000       0.276       0.638
                                    Roots                                    
=============================================================================
                  Real          Imaginary           Modulus         Frequency
-----------------------------------------------------------------------------
AR.1            2.1877           +0.0000j            2.1877            0.0000
-----------------------------------------------------------------------------

	county	totalcountconfirmed	totalcountdeaths	newcountconfirmed	newcountdeaths	date	NEVER	RARELY	SOMETIMES	FREQUENTLY	ALWAYS	population	confirmedper100k	newlyconfirmedper100k	deathper100k	newdeath100k	7dayrollingavg_newlyconfirmed	7dayrollingavg_newdeath	7dayrollingavg_newlyconfirmed_rawnumber	7dayrollingavg_newdeath_rawnumber
0	Alameda	29.0	0.0	29	0	2020-03-18	0.019	0.008	0.055	0.123	0.795	1671329	1.735146	1.735146	0.000000	0.000000	NaN	NaN	NaN	NaN
1	Alameda	36.0	0.0	7	0	2020-03-19	0.019	0.008	0.055	0.123	0.795	1671329	2.153974	0.418828	0.000000	0.000000	NaN	NaN	NaN	NaN
2	Alameda	42.0	0.0	6	0	2020-03-20	0.019	0.008	0.055	0.123	0.795	1671329	2.512970	0.358996	0.000000	0.000000	NaN	NaN	NaN	NaN
3	Alameda	62.0	0.0	20	0	2020-03-21	0.019	0.008	0.055	0.123	0.795	1671329	3.709623	1.196652	0.000000	0.000000	NaN	NaN	NaN	NaN
4	Alameda	72.0	1.0	10	1	2020-03-22	0.019	0.008	0.055	0.123	0.795	1671329	4.307949	0.598326	0.059833	0.059833	NaN	NaN	NaN	NaN

	totalcountconfirmed	totalcountdeaths	newcountconfirmed	newcountdeaths	NEVER	RARELY	SOMETIMES	FREQUENTLY	ALWAYS	population	confirmedper100k	newlyconfirmedper100k	deathper100k	newdeath100k	7dayrollingavg_newlyconfirmed	7dayrollingavg_newdeath	7dayrollingavg_newlyconfirmed_rawnumber	7dayrollingavg_newdeath_rawnumber
count	1.812400e+04	18124.000000	18124.000000	18124.000000	18124.000000	18124.000000	18124.000000	18124.000000	18124.000000	1.812400e+04	18124.000000	18124.000000	18124.000000	18124.000000	17776.000000	17776.000000	17776.000000	17776.000000
mean	1.327444e+04	210.774443	173.013187	2.048003	0.032279	0.030699	0.063650	0.158545	0.714792	6.823606e+05	1542.530785	20.284986	19.606417	0.210988	20.232502	0.208192	172.290086	2.001583
std	5.201773e+04	852.430551	814.985729	11.017366	0.028533	0.026178	0.033898	0.040410	0.092762	1.456208e+06	2084.332897	43.944282	28.624650	0.831238	31.417552	0.404034	775.800926	9.438015
min	0.000000e+00	0.000000	-1157.000000	-16.000000	0.001000	0.000000	0.004000	0.058000	0.482000	1.129000e+03	0.000000	-74.546183	0.000000	-5.543545	-6.977150	-0.791935	-1.142857	-1.428571
25%	9.400000e+01	1.000000	1.000000	0.000000	0.015000	0.013000	0.040000	0.134000	0.661000	4.590500e+04	99.646385	0.606875	1.002456	0.000000	2.049615	0.000000	2.000000	0.000000
50%	1.044000e+03	15.000000	14.000000	0.000000	0.023000	0.023000	0.058000	0.156000	0.737000	1.928430e+05	711.532789	6.700781	8.138623	0.000000	8.310360	0.059833	18.285714	0.142857
75%	8.147250e+03	114.000000	85.000000	0.000000	0.043000	0.043000	0.084000	0.186000	0.786000	7.621480e+05	2248.121154	22.165620	27.829294	0.000000	24.251025	0.230109	89.464286	1.000000
max	1.048757e+06	15260.000000	28549.000000	318.000000	0.140000	0.135000	0.162000	0.276000	0.889000	1.003911e+07	15791.711641	1860.053144	272.052534	27.717723	566.793296	6.335479	15711.142857	241.285714

	county	date	7dayrollingavg_newlyconfirmed	7dayrollingavg_newdeath
0	Alameda	2020-03-18	NaN	NaN
1	Alameda	2020-03-19	NaN	NaN
2	Alameda	2020-03-20	NaN	NaN
3	Alameda	2020-03-21	NaN	NaN
4	Alameda	2020-03-22	NaN	NaN

	county	todays_date	hospitalized_covid_patients	all_hospital_beds	icu_available_beds
0	Plumas	2020-03-29	1.0	NaN	NaN
1	Tehama	2020-03-29	0.0	308.0	2.0
2	Glenn	2020-03-29	0.0	NaN	NaN
3	Mono	2020-03-29	1.0	308.0	2.0
4	Marin	2020-03-29	20.0	493.0	11.0

	county	date	7dayrollingavg_newlyconfirmed	7dayrollingavg_newdeath	todays_date	hospitalized_covid_patients	all_hospital_beds	icu_available_beds
0	Alameda	2020-03-29	1.153915	0.042738	2020-03-29	127.0	2725.0	119.0
1	Alameda	2020-03-30	1.384698	0.051285	2020-03-30	148.0	3283.0	146.0
2	Alameda	2020-03-31	1.478721	0.042738	2020-03-31	150.0	1857.0	77.0
3	Alameda	2020-04-01	1.307770	0.042738	2020-04-01	128.0	2456.0	106.0
4	Alameda	2020-04-02	1.478721	0.051285	2020-04-02	133.0	2580.0	112.0

3) Modeling¶

Initial LSTM Model¶

ARIMA model¶

Modeling for 7 Counties Highest New Cases/100k¶

Tehama County¶

Riverside County¶

	7dayrollingavg_newlyconfirmed	7dayrollingavg_newdeath	hospitalized_covid_patients	all_hospital_beds	icu_available_beds
date
2020-03-29	2.468916	0.048382	1621.0	7394.0	345.0
2020-03-30	2.957007	0.056920	1878.0	9688.0	456.0
2020-03-31	3.369679	0.059766	1952.0	9460.0	445.0
2020-04-01	3.951689	0.079688	2071.0	10431.0	492.0
2020-04-02	4.219214	0.086803	2088.0	10349.0	488.0
...	...	...	...	...	...
2021-01-20	110.458032	2.033476	7413.0	19392.0	236.0
2021-01-21	97.337912	2.002170	7226.0	19411.0	274.0
2021-01-22	88.508157	1.999324	7014.0	19377.0	281.0
2021-01-23	82.928548	2.022092	6793.0	19334.0	299.0
2021-01-24	79.302728	2.009285	6606.0	19292.0	295.0

	hospitalized_covid_patients
2021-02-11	NaN
2021-02-12	NaN
2021-02-13	NaN
2021-02-14	NaN
2021-02-15	NaN

	Prediction
2021-01-25	133.826202
2021-01-26	163.729935
2021-01-27	169.911285
2021-01-28	171.605011
2021-01-29	172.200745

	hospitalized_covid_patients	7dayrollingavg_newdeath	all_hospital_beds	icu_available_beds
date
2021-01-02	7971.0	1.770220	20244.0	327.0
2021-01-03	8203.0	1.837101	20328.0	326.0
2021-01-04	8318.0	1.842793	20371.0	338.0
2021-01-05	8422.0	1.834255	20495.0	313.0
2021-01-06	8385.0	1.810064	20435.0	328.0

Dep. Variable:	y	No. Observations:	92
Model:	SARIMAX(0, 2, 2)	Log Likelihood	-283.304
Date:	Fri, 29 Jan 2021	AIC	572.608
Time:	17:42:58	BIC	580.108
Sample:	0	HQIC	575.632
	- 92
Covariance Type:	opg

	coef	std err	z	P>\|z\|	[0.025	0.975]
ma.L1	-1.1881	0.052	-22.821	0.000	-1.290	-1.086
ma.L2	0.2782	0.064	4.331	0.000	0.152	0.404
sigma2	30.9945	2.022	15.330	0.000	27.032	34.957

Ljung-Box (L1) (Q):	0.00	Jarque-Bera (JB):	298.43
Prob(Q):	0.97	Prob(JB):	0.00
Heteroskedasticity (H):	34.47	Skew:	0.23
Prob(H) (two-sided):	0.00	Kurtosis:	11.91

Ljung-Box (L1) (Q):	0.01	Jarque-Bera (JB):	353.54
Prob(Q):	0.93	Prob(JB):	0.00
Heteroskedasticity (H):	40.14	Skew:	0.76
Prob(H) (two-sided):	0.00	Kurtosis:	11.73

	coef	std err	z	P>\|z\|	[0.025	0.975]
ma.L1	-1.0460	0.050	-21.016	0.000	-1.144	-0.948
ma.L2	0.1922	0.065	2.938	0.003	0.064	0.320
sigma2	30.5668	1.908	16.017	0.000	26.827	34.307