109_.predicting_covid19_cases_using_pythonpy.py

#!/usr/bin/env python
__author__ = "Sreenivas Bhattiprolu"
__license__ = "Feel free to copy, I appreciate if you acknowledge Python for Microscopists"

# https://youtu.be/7sljAbwOhIw

#To download the file locally
#import urllib
#url = "https://covid.ourworldindata.org/data/ecdc/full_data.csv"
#urllib.request.urlretrieve (url, "data/full_data.csv")

import pandas as pd
import datetime as dt
from matplotlib import pyplot as plt
import matplotlib
import scipy
import numpy as np
import seaborn as sns


#date	location	new_cases	new_deaths	total_cases	total_deaths
CVD = pd.read_csv('https://covid.ourworldindata.org/data/ecdc/full_data.csv')
#print(CVD.head())
#print(CVD.dtypes)

#dateFormat = '%Y-%m-%d'
# Convert string values of date to datetime format
CVD['date'] = [dt.datetime.strptime(x,'%Y-%m-%d') for x in CVD['date']] 
#print(CVD.dtypes)
#Check for missing data
#print(CVD.isnull().sum()) #No missing data


#Change column titles to something appropriate
CVD.columns = ['Date', 'Country', 'New Cases', 'New deaths', 'Total Cases', 'Total Deaths' ]

#Select all countries except for china and World
CVD_no_china = CVD.loc[~(CVD['Country'].isin(["China", "World"]))]

#Group them by location and date, select only total cases and deaths for closer observation
#Reset index because groupby by default makes grouped columns indices
CVD_no_china = pd.DataFrame(CVD_no_china.groupby(['Country', 'Date'])['Total Cases', 'Total Deaths'].sum()).reset_index()
#print(CVD_no_china)

#Sort values by each country and by date - descending. Easy to interpret plots
CVD_no_china = CVD_no_china.sort_values(by = ['Country','Date'], ascending=False)
#print(CVD_no_china)



#####################################
#PREDICTION

def plot_exponential_fit_data(d_df, title, delta):
    d_df = d_df.sort_values(by=['Date'], ascending=True)
    d_df['x'] = np.arange(len(d_df)) + 1  #Add column x to the dataframe 
    d_df['y'] = d_df['Total Cases']   #Add column y to the dataframe 

    x = d_df['x'][:-delta]  #Remove delta number of data points (so we can predict them)
    y = d_df['y'][:-delta]  #Remove delta number of data points (so we can predict them)

#Use non-linear least squares to fit a function, f, to data.
#Let us fit data to exponential function: #y = Ae^(Bt)
    
    c2 = scipy.optimize.curve_fit(lambda t, a, b: a*np.exp(b*t),  x,  y,  p0=(20, 0.2)) 
# Function: lambda t, a, b: a*np.exp(b*t)
# xm y and po for initial values. 
    
    A, B = c2[0]  #Coefficients
    print(f'(y = Ae^(Bx)) A: {A}, B: {B}\n')
    x = range(1,d_df.shape[0] + 1)
    y_fit = A * np.exp(B * x)
#    print(y_fit)
    f, ax = plt.subplots(1,1, figsize=(12,6))
    g = sns.scatterplot(x=d_df['x'][:-delta], y=d_df['y'][:-delta], label='Confirmed cases (used for model creation)', color='red')
    g = sns.scatterplot(x=d_df['x'][-delta:], y=d_df['y'][-delta:], label='Confirmed cases (not used for model, va;idation)', color='blue')
    g = sns.lineplot(x=x, y=y_fit, label='Predicted values', color='green')  #Predicted
    x_future=range(85,90) #As of 24 March 2020 we have 85 days of info. 
    y_future=A * np.exp(B * x_future)
    print("Expected cases for the next 5 days: \n", y_future)
    plt.xlabel('Days since first case')
    plt.ylabel(f'Total cases')
    plt.title(f'Confirmed cases & projected cases: {title}')
    plt.xticks(rotation=90)
    ax.grid(color='black', linestyle='dotted', linewidth=0.75)
    plt.show()
    

CVD_USA = CVD_no_china[CVD_no_china['Country']=='United States']
CVD_Italy = CVD_no_china[CVD_no_china['Country']=='Italy']
CVD_South_Korea = CVD_no_china[CVD_no_china['Country']=='South Korea']

d_df = CVD_USA.copy()
plot_exponential_fit_data(d_df, 'USA', 5)