In [5]:
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
from matplotlib import ticker
import pycountry_convert as pc
import folium
import branca
from datetime import datetime, timedelta,date
from scipy.interpolate import make_interp_spline, BSpline
import plotly.express as px
import json, requests
import calmap
from keras.layers import Input, Dense, Activation, LeakyReLU
from keras import models
from keras.optimizers import RMSprop, Adam
import warnings
warnings.filterwarnings('ignore')
%matplotlib inline
In [7]:
import tensorflow as tf
print("Num GPUs Available: ", len(tf.config.experimental.list_physical_devices('GPU')))
In [8]:
tf.debugging.set_log_device_placement(True)
# Create some tensors
a = tf.constant([[1.0, 2.0, 3.0], [4.0, 5.0, 6.0]])
b = tf.constant([[1.0, 2.0], [3.0, 4.0], [5.0, 6.0]])
c = tf.matmul(a, b)
print(c)
In [9]:
# retrieve Dataset
df_confirmed = 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')
df_deaths = 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')
# Deprecated
# df_recovered = pd.read_csv('https://raw.githubusercontent.com/CSSEGISandData/COVID-19/master/csse_covid_19_data/csse_covid_19_time_series/time_series_19-covid-Recovered.csv')
df_covid19 = pd.read_csv("https://raw.githubusercontent.com/CSSEGISandData/COVID-19/web-data/data/cases_country.csv")
df_table = pd.read_csv("https://raw.githubusercontent.com/CSSEGISandData/COVID-19/web-data/data/cases_time.csv",parse_dates=['Last_Update'])
In [10]:
# new dataset
df_covid19 = df_covid19.drop(["People_Tested","People_Hospitalized","UID","ISO3","Mortality_Rate"],axis =1)
df_covid19.head(2)
Out[10]:
In [11]:
df_confirmed.head(2)
Out[11]:
Preprocessing¶
In [12]:
df_confirmed = df_confirmed.rename(columns={"Province/State":"state","Country/Region": "country"})
df_deaths = df_deaths.rename(columns={"Province/State":"state","Country/Region": "country"})
df_covid19 = df_covid19.rename(columns={"Country_Region": "country"})
df_covid19["Active"] = df_covid19["Confirmed"]-df_covid19["Recovered"]-df_covid19["Deaths"]
# df_recovered = df_recovered.rename(columns={"Province/State":"state","Country/Region": "country"})
In [13]:
# Changing the conuntry names as required by pycountry_convert Lib
df_confirmed.loc[df_confirmed['country'] == "US", "country"] = "USA"
df_deaths.loc[df_deaths['country'] == "US", "country"] = "USA"
df_covid19.loc[df_covid19['country'] == "US", "country"] = "USA"
df_table.loc[df_table['Country_Region'] == "US", "Country_Region"] = "USA"
# df_recovered.loc[df_recovered['country'] == "US", "country"] = "USA"
df_confirmed.loc[df_confirmed['country'] == 'Korea, South', "country"] = 'South Korea'
df_deaths.loc[df_deaths['country'] == 'Korea, South', "country"] = 'South Korea'
df_covid19.loc[df_covid19['country'] == "Korea, South", "country"] = "South Korea"
df_table.loc[df_table['Country_Region'] == "Korea, South", "Country_Region"] = "South Korea"
# df_recovered.loc[df_recovered['country'] == 'Korea, South', "country"] = 'South Korea'
df_confirmed.loc[df_confirmed['country'] == 'Taiwan*', "country"] = 'Taiwan'
df_deaths.loc[df_deaths['country'] == 'Taiwan*', "country"] = 'Taiwan'
df_covid19.loc[df_covid19['country'] == "Taiwan*", "country"] = "Taiwan"
df_table.loc[df_table['Country_Region'] == "Taiwan*", "Country_Region"] = "Taiwan"
# df_recovered.loc[df_recovered['country'] == 'Taiwan*', "country"] = 'Taiwan'
df_confirmed.loc[df_confirmed['country'] == 'Congo (Kinshasa)', "country"] = 'Democratic Republic of the Congo'
df_deaths.loc[df_deaths['country'] == 'Congo (Kinshasa)', "country"] = 'Democratic Republic of the Congo'
df_covid19.loc[df_covid19['country'] == "Congo (Kinshasa)", "country"] = "Democratic Republic of the Congo"
df_table.loc[df_table['Country_Region'] == "Congo (Kinshasa)", "Country_Region"] = "Democratic Republic of the Congo"
# df_recovered.loc[df_recovered['country'] == 'Congo (Kinshasa)', "country"] = 'Democratic Republic of the Congo'
df_confirmed.loc[df_confirmed['country'] == "Cote d'Ivoire", "country"] = "Côte d'Ivoire"
df_deaths.loc[df_deaths['country'] == "Cote d'Ivoire", "country"] = "Côte d'Ivoire"
df_covid19.loc[df_covid19['country'] == "Cote d'Ivoire", "country"] = "Côte d'Ivoire"
df_table.loc[df_table['Country_Region'] == "Cote d'Ivoire", "Country_Region"] = "Côte d'Ivoire"
# df_recovered.loc[df_recovered['country'] == "Cote d'Ivoire", "country"] = "Côte d'Ivoire"
df_confirmed.loc[df_confirmed['country'] == "Reunion", "country"] = "Réunion"
df_deaths.loc[df_deaths['country'] == "Reunion", "country"] = "Réunion"
df_covid19.loc[df_covid19['country'] == "Reunion", "country"] = "Réunion"
df_table.loc[df_table['Country_Region'] == "Reunion", "Country_Region"] = "Réunion"
# df_recovered.loc[df_recovered['country'] == "Reunion", "country"] = "Réunion"
df_confirmed.loc[df_confirmed['country'] == 'Congo (Brazzaville)', "country"] = 'Republic of the Congo'
df_deaths.loc[df_deaths['country'] == 'Congo (Brazzaville)', "country"] = 'Republic of the Congo'
df_covid19.loc[df_covid19['country'] == "Congo (Brazzaville)", "country"] = "Republic of the Congo"
df_table.loc[df_table['Country_Region'] == "Congo (Brazzaville)", "Country_Region"] = "Republic of the Congo"
# df_recovered.loc[df_recovered['country'] == 'Congo (Brazzaville)', "country"] = 'Republic of the Congo'
df_confirmed.loc[df_confirmed['country'] == 'Bahamas, The', "country"] = 'Bahamas'
df_deaths.loc[df_deaths['country'] == 'Bahamas, The', "country"] = 'Bahamas'
df_covid19.loc[df_covid19['country'] == "Bahamas, The", "country"] = "Bahamas"
df_table.loc[df_table['Country_Region'] == "Bahamas, The", "Country_Region"] = "Bahamas"
# df_recovered.loc[df_recovered['country'] == 'Bahamas, The', "country"] = 'Bahamas'
df_confirmed.loc[df_confirmed['country'] == 'Gambia, The', "country"] = 'Gambia'
df_deaths.loc[df_deaths['country'] == 'Gambia, The', "country"] = 'Gambia'
df_covid19.loc[df_covid19['country'] == "Gambia, The", "country"] = "Gambia"
df_table.loc[df_table['Country_Region'] == "Gambia", "Country_Region"] = "Gambia"
# df_recovered.loc[df_recovered['country'] == 'Gambia, The', "country"] = 'Gambia'
# getting all countries
countries = np.asarray(df_confirmed["country"])
countries1 = np.asarray(df_covid19["country"])
# Continent_code to Continent_names
continents = {
'NA': 'North America',
'SA': 'South America',
'AS': 'Asia',
'OC': 'Australia',
'AF': 'Africa',
'EU' : 'Europe',
'na' : 'Others'
}
# Defininng Function for getting continent code for country.
def country_to_continent_code(country):
try:
return pc.country_alpha2_to_continent_code(pc.country_name_to_country_alpha2(country))
except :
return 'na'
#Collecting Continent Information
df_confirmed.insert(2,"continent", [continents[country_to_continent_code(country)] for country in countries[:]])
df_deaths.insert(2,"continent", [continents[country_to_continent_code(country)] for country in countries[:]])
df_covid19.insert(1,"continent", [continents[country_to_continent_code(country)] for country in countries1[:]])
df_table.insert(1,"continent", [continents[country_to_continent_code(country)] for country in df_table["Country_Region"].values])
# df_recovered.insert(2,"continent", [continents[country_to_continent_code(country)] for country in countries[:]] )
In [14]:
df_table = df_table[df_table["continent"] != "Others"]
In [15]:
df_deaths[df_deaths["continent" ]== 'Others']
Out[15]:
In [16]:
df_confirmed = df_confirmed.replace(np.nan, '', regex=True)
df_deaths = df_deaths.replace(np.nan, '', regex=True)
# df_recovered = df_recovered.replace(np.nan, '', regex=True)
# df_active = df_active.replace(np.nan, '', regex=True)
Defining Functions¶
- plot_params()
- visualize_covid_cases()
- get_mortality_rate()
In [17]:
def plot_params(ax,axis_label= None, plt_title = None,label_size=15, axis_fsize = 15, title_fsize = 20, scale = 'linear' ):
# Tick-Parameters
ax.xaxis.set_minor_locator(ticker.AutoMinorLocator())
ax.yaxis.set_minor_locator(ticker.AutoMinorLocator())
ax.tick_params(which='both', width=1,labelsize=label_size)
ax.tick_params(which='major', length=6)
ax.tick_params(which='minor', length=3, color='0.8')
# Grid
plt.grid(lw = 1, ls = '-', c = "0.7", which = 'major')
plt.grid(lw = 1, ls = '-', c = "0.9", which = 'minor')
# Plot Title
plt.title( plt_title,{'fontsize':title_fsize})
# Yaxis sacle
plt.yscale(scale)
plt.minorticks_on()
# Plot Axes Labels
xl = plt.xlabel(axis_label[0],fontsize = axis_fsize)
yl = plt.ylabel(axis_label[1],fontsize = axis_fsize)
def visualize_covid_cases(confirmed, deaths, continent=None , country = None , state = None, period = None, figure = None, scale = "linear"):
x = 0
if figure == None:
f = plt.figure(figsize=(10,10))
# Sub plot
ax = f.add_subplot(111)
else :
f = figure[0]
# Sub plot
ax = f.add_subplot(figure[1],figure[2],figure[3])
ax.set_axisbelow(True)
plt.tight_layout(pad=10, w_pad=5, h_pad=5)
stats = [confirmed, deaths]
label = ["Confirmed", "Deaths"]
if continent != None:
params = ["continent",continent]
elif country != None:
params = ["country",country]
else:
params = ["All", "All"]
color = ["darkcyan","crimson"]
marker_style = dict(linewidth=3, linestyle='-', marker='o',markersize=4, markerfacecolor='#ffffff')
for i,stat in enumerate(stats):
if params[1] == "All" :
cases = np.sum(np.asarray(stat.iloc[:,5:]),axis = 0)[x:]
else :
cases = np.sum(np.asarray(stat[stat[params[0]] == params[1]].iloc[:,5:]),axis = 0)[x:]
date = np.arange(1,cases.shape[0]+1)[x:]
plt.plot(date,cases,label = label[i]+" (Total : "+str(cases[-1])+")",color=color[i],**marker_style)
plt.fill_between(date,cases,color=color[i],alpha=0.3)
if params[1] == "All" :
Total_confirmed = np.sum(np.asarray(stats[0].iloc[:,5:]),axis = 0)[x:]
Total_deaths = np.sum(np.asarray(stats[1].iloc[:,5:]),axis = 0)[x:]
else :
Total_confirmed = np.sum(np.asarray(stats[0][stat[params[0]] == params[1]].iloc[:,5:]),axis = 0)[x:]
Total_deaths = np.sum(np.asarray(stats[1][stat[params[0]] == params[1]].iloc[:,5:]),axis = 0)[x:]
text = "From "+stats[0].columns[5]+" to "+stats[0].columns[-1]+"\n"
text += "Mortality rate : "+ str(int(Total_deaths[-1]/(Total_confirmed[-1])*10000)/100)+"\n"
text += "Last 5 Days:\n"
text += "Confirmed : " + str(Total_confirmed[-1] - Total_confirmed[-6])+"\n"
text += "Deaths : " + str(Total_deaths[-1] - Total_deaths[-6])+"\n"
text += "Last 24 Hours:\n"
text += "Confirmed : " + str(Total_confirmed[-1] - Total_confirmed[-2])+"\n"
text += "Deaths : " + str(Total_deaths[-1] - Total_deaths[-2])+"\n"
plt.text(0.02, 0.78, text, fontsize=15, horizontalalignment='left', verticalalignment='top', transform=ax.transAxes,bbox=dict(facecolor='white', alpha=0.4))
# Plot Axes Labels
axis_label = ["Days ("+df_confirmed.columns[5]+" - "+df_confirmed.columns[-1]+")","No of Cases"]
# Plot Parameters
plot_params(ax,axis_label,scale = scale)
# Plot Title
if params[1] == "All" :
plt.title("COVID-19 Cases World",{'fontsize':25})
else:
plt.title("COVID-19 Cases for "+params[1] ,{'fontsize':25})
# Legend Location
l = plt.legend(loc= "best",fontsize = 15)
if figure == None:
plt.show()
def get_total_cases(cases, country = "All"):
if(country == "All") :
return np.sum(np.asarray(cases.iloc[:,5:]),axis = 0)[-1]
else :
return np.sum(np.asarray(cases[cases["country"] == country].iloc[:,5:]),axis = 0)[-1]
def get_mortality_rate(confirmed,deaths, continent = None, country = None):
if continent != None:
params = ["continent",continent]
elif country != None:
params = ["country",country]
else :
params = ["All", "All"]
if params[1] == "All" :
Total_confirmed = np.sum(np.asarray(confirmed.iloc[:,5:]),axis = 0)
Total_deaths = np.sum(np.asarray(deaths.iloc[:,5:]),axis = 0)
mortality_rate = np.round((Total_deaths/(Total_confirmed+1.01))*100,2)
else :
Total_confirmed = np.sum(np.asarray(confirmed[confirmed[params[0]] == params[1]].iloc[:,5:]),axis = 0)
Total_deaths = np.sum(np.asarray(deaths[deaths[params[0]] == params[1]].iloc[:,5:]),axis = 0)
mortality_rate = np.round((Total_deaths/(Total_confirmed+1.01))*100,2)
return np.nan_to_num(mortality_rate)
def dd(date1,date2):
return (datetime.strptime(date1,'%m/%d/%y') - datetime.strptime(date2,'%m/%d/%y')).days
out = ""#+"output/"
Exploratory Analysis of Data¶
In [18]:
df_countries_cases = df_covid19.copy().drop(['Lat','Long_','continent','Last_Update'],axis =1)
df_countries_cases.index = df_countries_cases["country"]
df_countries_cases = df_countries_cases.drop(['country'],axis=1)
df_continents_cases = df_covid19.copy().drop(['Lat','Long_','country','Last_Update'],axis =1)
df_continents_cases = df_continents_cases.groupby(["continent"]).sum()
df_countries_cases.fillna(0,inplace=True)
df_continents_cases.fillna(0,inplace=True)
Global Reported Cases as of May 9th, 2020¶
In [19]:
df_t = pd.DataFrame(pd.to_numeric(df_countries_cases.sum()),dtype=np.float64).transpose()
df_t["Mortality Rate (per 100)"] = np.round(100*df_t["Deaths"]/df_t["Confirmed"],2)
df_t.style.background_gradient(cmap='Wistia',axis=1).format("{:.0f}",subset=["Confirmed"])
Out[19]:
Separate by Continent¶
In [20]:
df_continents_cases["Mortality Rate (per 100)"] = np.round(100*df_continents_cases["Deaths"]/df_continents_cases["Confirmed"],2)
df_continents_cases.style.background_gradient(cmap='Blues',subset=["Confirmed"])\
.background_gradient(cmap='Reds',subset=["Deaths"])\
.background_gradient(cmap='Greens',subset=["Recovered"])\
.background_gradient(cmap='Purples',subset=["Active"])\
.background_gradient(cmap='Pastel1_r',subset=["Incident_Rate"])\
.background_gradient(cmap='YlOrBr',subset=["Mortality Rate (per 100)"])
Out[20]:
Separate by Country¶
In [21]:
# df_countries_cases.sort_values('Confirmed', ascending= False).style.background_gradient(cmap='Wistia')
df_countries_cases["Mortality Rate (per 100)"] = np.round(100*df_countries_cases["Deaths"]/df_countries_cases["Confirmed"],2)
df_countries_cases.sort_values('Confirmed', ascending= False).style.background_gradient(cmap='Blues',subset=["Confirmed"])\
.background_gradient(cmap='Reds',subset=["Deaths"])\
.background_gradient(cmap='Greens',subset=["Recovered"])\
.background_gradient(cmap='Purples',subset=["Active"])\
.background_gradient(cmap='Pastel1_r',subset=["Incident_Rate"])\
.background_gradient(cmap='YlOrBr',subset=["Mortality Rate (per 100)"])
Out[21]:
Top 10 Countries by confirmed cases and deaths¶
In [22]:
f = plt.figure(figsize=(10,5))
f.add_subplot(111)
plt.axes(axisbelow=True)
plt.barh(df_countries_cases.sort_values('Confirmed')["Confirmed"].index[-10:],df_countries_cases.sort_values('Confirmed')["Confirmed"].values[-10:],color="darkcyan")
plt.tick_params(size=5,labelsize = 13)
plt.xlabel("Confirmed Cases",fontsize=18)
plt.title("Top 10 Countries (Confirmed Cases)",fontsize=20)
plt.grid(alpha=0.3)
plt.savefig(out+'Top 10 Countries (Confirmed Cases).png')
In [23]:
f = plt.figure(figsize=(10,5))
f.add_subplot(111)
plt.axes(axisbelow=True)
plt.barh(df_countries_cases.sort_values('Deaths')["Deaths"].index[-10:],df_countries_cases.sort_values('Deaths')["Deaths"].values[-10:],color="crimson")
plt.tick_params(size=5,labelsize = 13)
plt.xlabel("Deaths Cases",fontsize=18)
plt.title("Top 10 Countries (Deaths Cases)",fontsize=20)
plt.grid(alpha=0.3,which='both')
plt.savefig(out+'Top 10 Countries (Deaths Cases).png')
In [24]:
f = plt.figure(figsize=(10,5))
f.add_subplot(111)
plt.axes(axisbelow=True)
plt.barh(df_countries_cases.sort_values('Active')["Active"].index[-10:],df_countries_cases.sort_values('Active')["Active"].values[-10:],color="purple")
plt.tick_params(size=5,labelsize = 13)
plt.xlabel("Active Cases",fontsize=18)
plt.title("Top 10 Countries (Active Cases)",fontsize=20)
plt.grid(alpha=0.3,which='both')
plt.savefig(out+'Top 10 Countries (Active Cases).png')
In [25]:
f = plt.figure(figsize=(10,5))
f.add_subplot(111)
plt.axes(axisbelow=True)
plt.barh(df_countries_cases.sort_values('Recovered')["Recovered"].index[-10:],df_countries_cases.sort_values('Recovered')["Recovered"].values[-10:],color="green")
plt.tick_params(size=5,labelsize = 13)
plt.xlabel("Recovered Cases",fontsize=18)
plt.title("Top 10 Countries (Recovered Cases)",fontsize=20)
plt.grid(alpha=0.3,which='both')
plt.savefig(out+'Top 10 Countries (Recovered Cases).png')
Correlation Analysis via Heat maps¶
In [26]:
df_continents_cases.iloc[:,:].corr().style.background_gradient(cmap='Reds')
Out[26]:
In [27]:
df_countries_cases.iloc[:,:].corr().style.background_gradient(cmap='Reds')
Out[27]:
Map Visualization¶
In [28]:
world_map = folium.Map(location=[10,0], tiles="cartodbpositron", zoom_start=2,max_zoom=6,min_zoom=2)
for i in range(0,len(df_confirmed)):
folium.Circle(
location=[df_confirmed.iloc[i]['Lat'], df_confirmed.iloc[i]['Long']],
tooltip = "<h5 style='text-align:center;font-weight: bold'>"+df_confirmed.iloc[i]['country']+"</h5>"+
"<div style='text-align:center;'>"+str(np.nan_to_num(df_confirmed.iloc[i]['state']))+"</div>"+
"<hr style='margin:10px;'>"+
"<ul style='color: #444;list-style-type:circle;align-item:left;padding-left:20px;padding-right:20px'>"+
"<li>Confirmed: "+str(df_confirmed.iloc[i,-1])+"</li>"+
"<li>Deaths: "+str(df_deaths.iloc[i,-1])+"</li>"+
"<li>Mortality Rate: "+str(np.round(df_deaths.iloc[i,-1]/(df_confirmed.iloc[i,-1]+1.00001)*100,2))+"</li>"+
"</ul>"
,
radius=(int((np.log(df_confirmed.iloc[i,-1]+1.00001)))+0.2)*50000,
color='#ff6600',
fill_color='#ff8533',
fill=True).add_to(world_map)
world_map
Out[28]:
In [29]:
temp_df = pd.DataFrame(df_countries_cases['Confirmed'])
temp_df = temp_df.reset_index()
fig = px.choropleth(temp_df, locations="country",
color=np.log10(temp_df["Confirmed"]), # lifeExp is a column of gapminder
hover_name="country", # column to add to hover information
hover_data=["Confirmed"],
color_continuous_scale=px.colors.sequential.Plasma,locationmode="country names")
fig.update_geos(fitbounds="locations", visible=False)
fig.update_layout(title_text="Confirmed Cases Heat Map (Log Scale)")
fig.update_coloraxes(colorbar_title="Confirmed Cases(Log Scale)",colorscale="Blues")
# fig.to_image("Global Heat Map confirmed.png")
fig.show()
In [30]:
temp_df = pd.DataFrame(df_countries_cases['Deaths'])
temp_df = temp_df.reset_index()
fig = px.choropleth(temp_df, locations="country",
color=np.log10(temp_df["Deaths"]+1), # lifeExp is a column of gapminder
hover_name="country", # column to add to hover information
hover_data=["Deaths"],
color_continuous_scale=px.colors.sequential.Plasma,locationmode="country names")
fig.update_geos(fitbounds="locations", visible=False)
fig.update_layout(title_text="Deaths Heat Map (Log Scale)")
fig.update_coloraxes(colorbar_title="Deaths (Log Scale)",colorscale="Reds")
# fig.to_image("Global Heat Map deaths.png")
fig.show()
In [31]:
temp_df = pd.DataFrame(df_countries_cases['Recovered'])
temp_df = temp_df.reset_index()
fig = px.choropleth(temp_df, locations="country",
color=np.log10(temp_df.iloc[:,-1]+1), # lifeExp is a column of gapminder
hover_name="country", # column to add to hover information
hover_data=["Recovered"],
color_continuous_scale=px.colors.sequential.Plasma,locationmode="country names")
fig.update_geos(fitbounds="locations", visible=False)
fig.update_layout(title_text="Recovered Heat Map (Log Scale)")
fig.update_coloraxes(colorbar_title="Recovered",colorscale="Greens")
# fig.to_image("Global Heat Map deaths.png")
fig.show()
In [32]:
df_data = df_table.groupby(['Last_Update', 'Country_Region'])['Confirmed', 'Deaths'].max().reset_index().fillna(0)
df_data["Last_Update"] = pd.to_datetime( df_data["Last_Update"]).dt.strftime('%m/%d/%Y')
fig = px.scatter_geo(df_data, locations="Country_Region", locationmode='country names',
color=np.power(df_data["Confirmed"],0.3)-2 , size= np.power(df_data["Confirmed"]+1,0.3)-1, hover_name="Country_Region",
hover_data=["Confirmed"],
range_color= [0, max(np.power(df_data["Confirmed"],0.3))],
projection="natural earth", animation_frame="Last_Update",
color_continuous_scale=px.colors.sequential.Plasma,
title='COVID-19: Progression of spread'
)
fig.update_coloraxes(colorscale="hot")
fig.update(layout_coloraxis_showscale=False)
fig.show()
COVID-19 Spread Analysis¶
- Spread across the world
- Spread trends for the world, continents, and countries
In [33]:
case_nums_country = df_confirmed.groupby("country").sum().drop(['Lat','Long'],axis =1).apply(lambda x: x[x > 0].count(), axis =0)
d = [datetime.strptime(date,'%m/%d/%y').strftime("%d %b") for date in case_nums_country.index]
f = plt.figure(figsize=(15,8))
f.add_subplot(111)
marker_style = dict(c="darkblue",linewidth=6, linestyle='-', marker='o',markersize=8, markerfacecolor='#ffffff')
plt.plot(d, case_nums_country,**marker_style)
plt.tick_params(labelsize = 14)
plt.xticks(list(np.arange(0,len(d),int(len(d)/5))),d[:-1:int(len(d)/5)]+[d[-1]])
#labels
plt.xlabel("Dates",fontsize=18)
plt.ylabel("Number of Countries/Regions",fontsize=1)
plt.grid(alpha = 0.3)
plt.savefig(out+'spread.png')
plt.show()
plt.close()
In [34]:
cols = 1
rows = 1
f = plt.figure(figsize=(10,10*rows))
visualize_covid_cases(df_confirmed, df_deaths,continent = "All",figure = [f,rows,cols, 1])
plt.savefig(out+'COIVD-19-World.png')
plt.show()
In [35]:
df_continents= df_confirmed.groupby(["continent"]).sum()
continents = df_continents.sort_values(df_continents.columns[-1],ascending = False).index
cols =2
rows = int(np.ceil(continents.shape[0]/cols))
f = plt.figure(figsize=(20,9*rows))
for i,continent in enumerate(continents):
visualize_covid_cases(df_confirmed, df_deaths, continent = continent,figure = [f,rows,cols, i+1])
plt.show()
In [36]:
df_countries = df_confirmed.groupby(["country"]).sum()
df_countries = df_countries.sort_values(df_countries.columns[-1],ascending = False)
countries = df_countries[df_countries[df_countries.columns[-1]] >= 4000].index
cols =2
rows = int(np.ceil(countries.shape[0]/cols))
f = plt.figure(figsize=(18,7*rows))
for i,country in enumerate(countries):
visualize_covid_cases(df_confirmed, df_deaths,country = country,figure = [f,rows,cols, i+1])
plt.show()
In [37]:
temp = df_confirmed.groupby('country').sum().drop(["Lat","Long"],axis =1).sort_values(df_confirmed.columns[-1], ascending= False)
threshold = 50
f = plt.figure(figsize=(15,12))
ax = f.add_subplot(111)
for i,country in enumerate(temp.index):
if i >= 9:
if country != "USA" and country != "Japan" :
continue
days = 60
t = temp.loc[temp.index== country].values[0]
t = t[t>threshold][:days]
date = np.arange(0,len(t[:days]))
xnew = np.linspace(date.min(), date.max(), 30)
spl = make_interp_spline(date, t, k=1) # type: BSpline
power_smooth = spl(xnew)
if country != "USA":
plt.plot(xnew,power_smooth,'-o',label = country,linewidth =3, markevery=[-1])
else:
marker_style = dict(linewidth=4, linestyle='-', marker='o',markersize=10, markerfacecolor='#ffffff')
plt.plot(date,t,"-.",label = country,**marker_style)
plt.tick_params(labelsize = 14)
plt.xticks(np.arange(0,days,7),[ "Day "+str(i) for i in range(days)][::7])
# Reference lines
x = np.arange(0,18)
y = 2**(x+np.log2(threshold))
plt.plot(x,y,"--",linewidth =2,color = "gray")
plt.annotate("No. of cases doubles every day",(x[-2],y[-1]),xycoords="data",fontsize=14,alpha = 0.5)
x = np.arange(0,int(days-22))
y = 2**(x/2+np.log2(threshold))
plt.plot(x,y,"--",linewidth =2,color = "gray")
plt.annotate(".. every second day",(x[-3],y[-1]),xycoords="data",fontsize=14,alpha = 0.5)
x = np.arange(0,int(days-4))
y = 2**(x/7+np.log2(threshold))
plt.plot(x,y,"--",linewidth =2,color = "gray")
plt.annotate(".. every week",(x[-3],y[-1]),xycoords="data",fontsize=14,alpha = 0.5)
x = np.arange(0,int(days-4))
y = 2**(x/30+np.log2(threshold))
plt.plot(x,y,"--",linewidth =2,color = "gray")
plt.annotate(".. every month",(x[-3],y[-1]),xycoords="data",fontsize=14,alpha = 0.5)
# India is following trend similar to doulbe the cases in 4 days but it may increase the rate
x = np.arange(0,int(days-5))
y = 2**(x/4+np.log2(threshold))
plt.plot(x,y,"--",linewidth =2,color = "Red")
plt.annotate(".. every 4 days",(x[-3],y[-1]),color="Red",xycoords="data",fontsize=14,alpha = 0.8)
# plot Params
plt.xlabel("Days",fontsize=17)
plt.ylabel("Number of Confirmed Cases",fontsize=17)
plt.title("Trend Comparison of Different Countries\n and USA (confirmed) ",fontsize=22)
plt.legend(loc = "upper left")
plt.yscale("log")
plt.grid(which="both")
plt.savefig(out+'Trend Comparison with USA (confirmed).png')
plt.show()
In [38]:
temp = df_deaths.groupby('country').sum().drop(["Lat","Long"],axis =1).sort_values(df_deaths.columns[-1], ascending= False)
threshold = 10
f = plt.figure(figsize=(15,12))
ax = f.add_subplot(111)
for i,country in enumerate(temp.index):
if i >= 10:
break
days = 60
t = temp.loc[temp.index== country].values[0]
t = t[t>threshold][:days]
date = np.arange(0,len(t[:days]))
xnew = np.linspace(date.min(), date.max(), 30)
spl = make_interp_spline(date, t, k=1) # type: BSpline
power_smooth = spl(xnew)
plt.plot(xnew,power_smooth,'-o',label = country,linewidth =3, markevery=[-1])
plt.tick_params(labelsize = 14)
plt.xticks(np.arange(0,days,7),[ "Day "+str(i) for i in range(days)][::7])
# Reference lines
x = np.arange(0,18)
y = 2**(x+np.log2(threshold))
plt.plot(x,y,"--",linewidth =2,color = "gray")
plt.annotate("No. of cases doubles every day",(x[-2],y[-1]),xycoords="data",fontsize=14,alpha = 0.5)
x = np.arange(0,days-22)
y = 2**(x/2+np.log2(threshold))
plt.plot(x,y,"--",linewidth =2,color = "gray")
plt.annotate(".. every second day",(x[-3],y[-1]),xycoords="data",fontsize=14,alpha = 0.5)
x = np.arange(0,days-3)
y = 2**(x/7+np.log2(threshold))
plt.plot(x,y,"--",linewidth =2,color = "gray")
plt.annotate(".. every week",(x[-3],y[-1]),xycoords="data",fontsize=14,alpha = 0.5)
x = np.arange(0,days-3)
y = 2**(x/30+np.log2(threshold))
plt.plot(x,y,"--",linewidth =2,color = "gray")
plt.annotate(".. every month",(x[-3],y[-1]),xycoords="data",fontsize=14,alpha = 0.5)
# plot Params
plt.xlabel("Days",fontsize=17)
plt.ylabel("Number of Deaths",fontsize=17)
plt.title("Trend Comparison of Different Countries \n(Deaths)",fontsize=22)
plt.legend(loc = "upper left")
plt.yscale("log")
plt.grid(which="both")
plt.savefig(out+'Trend Comparison countries deaths.png')
plt.show()
In [39]:
temp = df_confirmed.groupby('continent').sum().drop(["Lat","Long"],axis =1).sort_values(df_confirmed.columns[-1], ascending= False)
threshold = 50
f = plt.figure(figsize=(15,12))
ax = f.add_subplot(111)
for i,country in enumerate(temp.index):
if i >= 10:
break
days = 60
t = temp.loc[temp.index== country].values[0]
t = t[t>threshold][:days]
date = np.arange(0,len(t[:days]))
xnew = np.linspace(date.min(), date.max(), 30)
spl = make_interp_spline(date, t, k=1) # type: BSpline
power_smooth = spl(xnew)
plt.plot(xnew,power_smooth,'-o',label = country,linewidth =3, markevery=[-1])
plt.tick_params(labelsize = 14)
plt.xticks(np.arange(0,days,7),[ "Day "+str(i) for i in range(days)][::7])
# Reference lines
x = np.arange(0,18)
y = 2**(x+np.log2(threshold))
plt.plot(x,y,"--",linewidth =2,color = "gray")
plt.annotate("No. of cases doubles every day",(x[-2],y[-1]),xycoords="data",fontsize=14,alpha = 0.5)
x = np.arange(0,days-22)
y = 2**(x/2+np.log2(threshold))
plt.plot(x,y,"--",linewidth =2,color = "gray")
plt.annotate(".. every second day",(x[-3],y[-1]),xycoords="data",fontsize=14,alpha = 0.5)
x = np.arange(0,days-2)
y = 2**(x/7+np.log2(threshold))
plt.plot(x,y,"--",linewidth =2,color = "gray")
plt.annotate(".. every week",(x[-3],y[-1]),xycoords="data",fontsize=14,alpha = 0.5)
x = np.arange(0,days-3)
y = 2**(x/30+np.log2(threshold))
plt.plot(x,y,"--",linewidth =2,color = "gray")
plt.annotate(".. every month",(x[-3],y[-1]),xycoords="data",fontsize=14,alpha = 0.5)
# plot Params
plt.xlabel("Days",fontsize=17)
plt.ylabel("Number of Confirmed Cases",fontsize=17)
plt.title("Trend Comparison of Different Continents \n(confirmed) ",fontsize=22)
plt.legend(loc = "upper left")
plt.yscale("log")
plt.grid(which="both")
plt.savefig(out+'Trend Comparison of continents (Confirmed).png')
plt.show()
In [40]:
temp = df_deaths.groupby('continent').sum().drop(["Lat","Long"],axis =1).sort_values(df_deaths.columns[-1], ascending= False)
threshold = 10
f = plt.figure(figsize=(15,12))
ax = f.add_subplot(111)
for i,country in enumerate(temp.index):
if i >= 10:
break
days = 60
t = temp.loc[temp.index== country].values[0]
t = t[t>threshold][:days]
date = np.arange(0,len(t[:days]))
xnew = np.linspace(date.min(), date.max(), 10)
spl = make_interp_spline(date, t, k=1) # type: BSpline
power_smooth = spl(xnew)
plt.plot(xnew,power_smooth,'-o',label = country,linewidth =3, markevery=[-1])
plt.tick_params(labelsize = 14)
plt.xticks(np.arange(0,days,7),[ "Day "+str(i) for i in range(days)][::7])
# Reference lines
x = np.arange(0,18)
y = 2**(x+np.log2(threshold))
plt.plot(x,y,"--",linewidth =2,color = "gray")
plt.annotate("No. of cases doubles every day",(x[-2],y[-1]),xycoords="data",fontsize=14,alpha = 0.5)
x = np.arange(0,days-22)
y = 2**(x/2+np.log2(threshold))
plt.plot(x,y,"--",linewidth =2,color = "gray")
plt.annotate(".. every second day",(x[-3],y[-1]),xycoords="data",fontsize=14,alpha = 0.5)
x = np.arange(0,days-3)
y = 2**(x/7+np.log2(threshold))
plt.plot(x,y,"--",linewidth =2,color = "gray")
plt.annotate(".. every week",(x[-3],y[-1]),xycoords="data",fontsize=14,alpha = 0.5)
x = np.arange(0,days-3)
y = 2**(x/30+np.log2(threshold))
plt.plot(x,y,"--",linewidth =2,color = "gray")
plt.annotate(".. every month",(x[-3],y[-1]),xycoords="data",fontsize=14,alpha = 0.5)
# plot Params
plt.xlabel("Days",fontsize=17)
plt.ylabel("Number of Deaths",fontsize=17)
plt.title("Trend Comparison of Different Continents \n(Deaths)",fontsize=22)
plt.legend(loc = "upper left")
plt.yscale("log")
plt.grid(which="both")
plt.savefig(out+'Trend Comparison continents (deaths).png')
plt.show()
In [41]:
temp_data = df_confirmed.iloc[:,5:].sum(axis =0)
f = plt.figure(figsize=(20,12))
f.add_subplot(111)
threshold = 100000
t = temp_data.values
t = t[t >threshold]
date = np.arange(0,len(t[:]))
xnew = np.linspace(date.min(), date.max(), 10)
spl = make_interp_spline(date, t, k=1) # type: BSpline
power_smooth = spl(xnew)
marker_style = dict(linewidth=4, linestyle='-', marker='o',markersize=10, markerfacecolor='#ffffff')
plt.plot(date,t,"-.",label="Confirmed Cases",**marker_style)
days = 80
# Reference lines
x = np.arange(0,days)
y = 2**(x+np.log2(threshold))
plt.plot(x,y,"--",linewidth =2,color = "gray")
plt.annotate("No. of Cases Doubles Every Day",(np.log2((t.max()-threshold)/threshold),t.max()-threshold/2),xycoords="data",fontsize=14,alpha = 0.5)
x = np.arange(0,days)
y = 2**(x/3+np.log2(threshold))
plt.plot(x,y,"--",linewidth =2,color = "gray")
plt.annotate("...Every Third Day",(np.log2((t.max()-threshold)/threshold)*3,t.max()-threshold),xycoords="data",fontsize=14,alpha = 0.5)
x = np.arange(0,days)
y = 2**(x/7+np.log2(threshold))
plt.plot(x,y,"--",linewidth =2,color = "gray")
plt.annotate("... Every Week",(np.log2((t.max()-threshold)/threshold)*7,t.max()-threshold),xycoords="data",fontsize=14,alpha = 0.5)
x = np.arange(0,days)
y = 2**(x/30+np.log2(threshold))
plt.plot(x,y,"--",linewidth =2,color = "gray")
plt.annotate(".. Every Month",(18,2**(17/30+np.log2(threshold))),xycoords="data",fontsize=14,alpha = 0.5)
x = np.arange(0,days)
y = 2**(x/11+np.log2(threshold))
plt.plot(x,y,"--",linewidth =2,color = "Red")
plt.annotate(".. Every 11 Days",(np.log2((t.max()-threshold)/threshold)*9,t.max()-threshold),color="Red",xycoords="data",fontsize=14,alpha = 0.8)
plt.xlim(date[0],date[-1])
plt.ylim(threshold - threshold/10,2*t.max())
# plot Params
# plt.tight_layout()
plt.tick_params(labelsize = 16)
plt.xticks(np.arange(0,len(t[:])+7,7),[ "Day "+str(i) for i in range(len(t[:])+7)][::7])
plt.xlabel("Days",fontsize=19)
plt.ylabel("Number of Confirmed Cases",fontsize=19)
plt.title("World Trend",fontsize=24)
plt.legend(loc = "upper left")
plt.yscale("log")
plt.grid(which="both")
plt.savefig(out+"World Trend Confirmed cases.png")
plt.show()
Machine Learning Model¶
In [42]:
Visible = Input(shape=(1,))
Dense_l1 = Dense(80,name="Dense_l1")(Visible)
LRelu_l1 = LeakyReLU(name = "LRelu_l1")(Dense_l1)
Dense_l2 = Dense(80,name = "Dense_l2")(LRelu_l1)
LRelu_l2 = LeakyReLU(name = "LRelu_l2")(Dense_l2)
Dense_l3 = Dense(80,name = "Dense_l3")(LRelu_l2)
LRelu_l3 = LeakyReLU(name = "LRelu_l3")(Dense_l3)
Dense_l4 = Dense(1,name="Dense_l4")(LRelu_l3)
LRelu_l4 = LeakyReLU(name = "Output")(Dense_l4
)
model = models.Model(inputs=Visible, outputs=LRelu_l4)
model.compile(optimizer=Adam(lr=0.001),
loss='mean_squared_error',
metrics=['accuracy'])
model.summary()
In [43]:
data_y = np.log10(np.asarray(df_confirmed.iloc[:,5:].sum(axis =0)).astype("float32"))
data_x = np.arange(1,len(data_y)+1)
In [44]:
model = models.load_model("model_confirmed_v3.h5")
model.summary()
In [50]:
M = 1000000
prediction_days = 10
temp_data = df_confirmed.iloc[:,5:].sum(axis =0)
data = np.power(10,model.predict(np.arange(1,len(temp_data)+prediction_days+1)))
f = plt.figure(figsize=(15,10))
ax = f.add_subplot(111)
date = np.arange(0,len(temp_data))
marker_style = dict(linewidth=3, linestyle='-', marker='o',markersize=7, markerfacecolor='#ffffff')
plt.plot(date,temp_data/M,"-.",color="darkcyan",**marker_style, label="Actual Curve")
date = np.arange(0,len(data))
plt.plot(date,data/M,"-.",color="orangered",label="Predicted Curve")
nextdays = [(datetime.strptime(d[-1],'%d %b')+timedelta(days=i)).strftime("%d %b") for i in range(1,prediction_days+1)]
total = d + nextdays
text = "Prediction for next "+str(prediction_days) +" days:\n"
for i in range(prediction_days):
text += nextdays[i]+" : "+str(np.round(data[-1*(prediction_days-i)],-3)[0]/M)+" M\n"
plt.text(0.02, 0.78, text, fontsize=17, horizontalalignment='left', verticalalignment='top', transform=ax.transAxes,bbox=dict(facecolor='white', alpha=0.4))
# X-axis
plt.xticks(list(np.arange(0,len(total),int(len(total)/5))),d[:-1:int(len(total)/5)]+[total[-1]])
# Tick-Parameters
ax.xaxis.set_minor_locator(ticker.AutoMinorLocator())
ax.yaxis.set_minor_locator(ticker.AutoMinorLocator())
ax.tick_params(which='both', width=1,labelsize=14)
ax.tick_params(which='major', length=6)
ax.tick_params(which='minor', length=3, color='0.8')
# Grid
plt.grid(lw = 1, ls = '-', c = "0.7", which = 'major')
plt.grid(lw = 1, ls = '-', c = "0.9", which = 'minor')
# Plot Title
plt.title("COVID-19 Next 10 day Prediction Curve-Global Confirmed Cases",{'fontsize':22})
# Axis Lable
plt.xlabel("Date",fontsize =18)
plt.ylabel("Number of Confirmed Cases (M)",fontsize =18)
# plt.yscale("log")
plt.legend(fontsize =18)
plt.tick_params(labelsize = 13)
plt.savefig(out+"Prediction Curve-Confirmed.png")
plt.show()
In [51]:
Visible = Input(shape=(1,))
Dense_l1 = Dense(80,name="Dense_l1")(Visible)
LRelu_l1 = LeakyReLU(name = "LRelu_l1")(Dense_l1)
Dense_l2 = Dense(80,name = "Dense_l2")(LRelu_l1)
LRelu_l2 = LeakyReLU(name = "LRelu_l2")(Dense_l2)
Dense_l3 = Dense(80,name = "Dense_l3")(LRelu_l2)
LRelu_l3 = LeakyReLU(name = "LRelu_l3")(Dense_l3)
Dense_l4 = Dense(1,name="Dense_l4")(LRelu_l3)
LRelu_l4 = LeakyReLU(name = "Output")(Dense_l4
)
model2 = models.Model(inputs=Visible, outputs=LRelu_l4)
model2.compile(optimizer=Adam(lr=0.001),
loss='mean_squared_error',
metrics=['accuracy'])
model2.summary()
In [52]:
data1_y = np.log10(np.asarray(df_deaths.iloc[:,5:].sum(axis =0)).astype("float32"))
data1_x = np.arange(1,len(data1_y)+1)
In [53]:
model2 = models.load_model("model_deaths_v3.h5")
model2.summary()
In [54]:
thousand = 1000
prediction_days = 10
temp_data = df_deaths.iloc[:,5:].sum(axis =0)
data = np.power(10,model2.predict(np.arange(1,len(temp_data)+prediction_days+1)))
f = plt.figure(figsize=(15,10))
ax = f.add_subplot(111)
date = np.arange(0,len(temp_data))
# xnew = np.linspace(date.min(), date.max(), 10)
# spl = make_interp_spline(date, t, k=1) # type: BSpline
# power_smooth = spl(xnew)
marker_style = dict(linewidth=3, linestyle='-', marker='o',markersize=7, markerfacecolor='#ffffff')
plt.plot(date,temp_data/thousand,"-.",color="maroon",**marker_style, label="Actual Curve")
date = np.arange(0,len(data))
plt.plot(date,data/thousand,"-.",color="orangered",label="Predicted Curve")
nextdays = [(datetime.strptime(d[-1],'%d %b')+timedelta(days=i)).strftime("%d %b") for i in range(1,prediction_days+1)]
total =d+nextdays
text = "Prediction for next "+str(prediction_days) +" days:\n"
for i in range(prediction_days):
text += nextdays[i]+" : "+str(np.round(data[-1*(prediction_days-i)],-1)[0]/thousand)+"K\n"
plt.text(0.02, 0.78, text, fontsize=17, horizontalalignment='left', verticalalignment='top', transform=ax.transAxes,bbox=dict(facecolor='white', alpha=0.4))
# X-axis
plt.xticks(list(np.arange(0,len(total),int(len(total)/5))),total[:-1:int(len(total)/5)]+[total[-1]])
# Tick-Parameters
ax.xaxis.set_minor_locator(ticker.AutoMinorLocator())
ax.yaxis.set_minor_locator(ticker.AutoMinorLocator())
ax.tick_params(which='both', width=1,labelsize=12)
ax.tick_params(which='major', length=6)
ax.tick_params(which='minor', length=3, color='0.8')
# Grid
plt.grid(lw = 1, ls = '-', c = "0.7", which = 'major')
plt.grid(lw = 1, ls = '-', c = "0.9", which = 'minor')
# Plot Title
plt.title("COVID-19 Next 10 day Prediction Curve-Global Deaths Cases",{'fontsize':22})
# Axis Lable
plt.xlabel("Date",fontsize =18)
plt.ylabel("Number of Deaths Cases (Thousand)",fontsize =18)
# plt.yscale("log")
plt.legend(fontsize =18)
plt.tick_params(labelsize = 13)
plt.savefig(out+"Prediction Curve2.png")
plt.show()
In [55]:
start_date = "4/5/20"
sd= dd(start_date,"1/22/20")
nextdays = [(datetime.strptime(d[-1],'%d %b')+timedelta(days=i)).strftime("%d %b") for i in range(1,prediction_days+1)]
total =d+nextdays
data_confirmed = np.power(10,model.predict(np.arange(1,len(data_y)+prediction_days+1)))
data_deaths = np.power(10,model2.predict(np.arange(1,len(data_y)+prediction_days+1)))
pd.DataFrame([total[sd:],
list(np.int64(np.round(data_confirmed[sd:].reshape(-1)/lakh,2)*lakh)),
list(df_confirmed.iloc[:,5:].sum(axis =0)[sd:]),
list(np.int64(np.round(data_deaths[sd:].reshape(-1)/thousand,2)*thousand)),
list(df_deaths.iloc[:,5:].sum(axis =0)[sd:])],
["Date","Confirmed(Predicted)","Confirmed(Actual)","Deaths(Predicted)","Deaths(Actual)"]
).transpose().set_index("Date")
Out[55]:
In [57]:
# Data
temp_data = np.nan_to_num(df_confirmed.sum()[5:].diff())
# Plot
f = plt.figure(figsize=(15,10))
ax = f.add_subplot(111)
date = np.arange(0,len(temp_data))
marker_style = dict(linewidth=2, linestyle='-', marker='o',markersize=5)
plt.plot(date,temp_data/thousand,"-.",color="blue",**marker_style)
nextdays = [(datetime.strptime(d[-1],'%d %b')+timedelta(days=i)).strftime("%d %b") for i in range(1,prediction_days+1)]
total =d+nextdays
# X-axis
plt.xticks(list(np.arange(0,len(total),int(len(total)/5))),total[:-1:int(len(total)/5)]+[total[-1]])
# Tick-Parameters
ax.xaxis.set_minor_locator(ticker.AutoMinorLocator())
ax.yaxis.set_minor_locator(ticker.AutoMinorLocator())
ax.tick_params(which='both', width=1,labelsize=12)
ax.tick_params(which='major', length=6)
ax.tick_params(which='minor', length=3, color='0.8')
# Grid
plt.grid(lw = 1, ls = '-', c = "0.85", which = 'major')
plt.grid(lw = 1, ls = '-', c = "0.95", which = 'minor')
# Plot Title
plt.title("COVID-19 Global Daily New Confirmed Cases",{'fontsize':22})
# Axis Lable
plt.xlabel("Date",fontsize =18)
plt.ylabel("Number of Daily Confirmed Cases (Thousand)",fontsize =18)
# plt.yscale("log")
plt.tick_params(labelsize = 13)
plt.savefig(out+"daily confirmed cases global.png")
plt.show()
In [58]:
# Data
temp_data = np.nan_to_num(df_deaths.sum()[5:].diff())
# Plot
f = plt.figure(figsize=(15,10))
ax = f.add_subplot(111)
date = np.arange(0,len(temp_data))
marker_style = dict(linewidth=2, linestyle='-', marker='o',markersize=5)
plt.plot(date,temp_data/thousand,"-.",color="red",**marker_style)
nextdays = [(datetime.strptime(d[-1],'%d %b')+timedelta(days=i)).strftime("%d %b") for i in range(1,prediction_days+1)]
total =d+nextdays
# X-axis
plt.xticks(list(np.arange(0,len(total),int(len(total)/5))),total[:-1:int(len(total)/5)]+[total[-1]])
# Tick-Parameters
ax.xaxis.set_minor_locator(ticker.AutoMinorLocator())
ax.yaxis.set_minor_locator(ticker.AutoMinorLocator())
ax.tick_params(which='both', width=1,labelsize=14)
ax.tick_params(which='major', length=6)
ax.tick_params(which='minor', length=3, color='0.8')
# Grid
plt.grid(lw = 1, ls = '-', c = "0.85", which = 'major')
plt.grid(lw = 1, ls = '-', c = "0.95", which = 'minor')
# Plot Title
plt.title("COVID-19 Global Daily Deaths Reported",{'fontsize':22})
# Axis Lable
plt.xlabel("Date",fontsize =18)
plt.ylabel("Number of Daily Deaths Reported (Thousand)",fontsize =18)
# plt.yscale("log")
plt.savefig(out+"daily deaths cases Global.png")
plt.show()
Testing¶
In [68]:
df_test = pd.read_csv("https://raw.githubusercontent.com/tarunk04/COVID-19-CaseStudy-and-Predictions/master/test_data_v70/test_data.csv")[:-1]
# df_test
In [69]:
df_test.drop(df_test[df_test["Entity"]=="India - people tested"].index[0],inplace=True)
df_test["country"] = [ c.split(" ")[:-3] for c in df_test["Entity"].values]
df_test = df_test.loc[:,["country","Cumulative total","Cumulative total per thousand"]]
In [70]:
df_test = df_test.sort_values("Cumulative total",ascending=False)[:23].reset_index().drop(['index'],axis=1)
df_test["country"] = [ c[0] for c in df_test["country"].values]
df_test.loc[0,"country"] ="USA"
df_test.loc[7,"country"] = "United Kingdom"
df_test.loc[12,"country"] = "South Korea"
df_test = df_test.drop([4,10,14])
df_test = df_test.reset_index().drop(['index'],axis=1)
df_test["Cumulative total per million"] = df_test["Cumulative total per thousand"]*1000
df_test = df_test.drop(["Cumulative total per thousand"],axis=1)
In [71]:
df_test["confirmed"] = [ df_countries_cases.loc[c,'Confirmed'] for c in df_test["country"].values]
df_test["deaths"] = [ df_countries_cases.loc[c,'Deaths'] for c in df_test["country"].values]
df_test["MR"] = np.round(100*df_test["deaths"]/df_test["confirmed"],2)
df_test["Positive"] = np.round(100*df_test["confirmed"]/df_test["Cumulative total"],2)
df_test.style.background_gradient(cmap='Blues',subset=["Cumulative total"])\
.background_gradient(cmap='Reds',subset=["Cumulative total per million"])\
.background_gradient(cmap='Greens',subset=["confirmed"])\
.background_gradient(cmap='Purples',subset=["deaths"])\
.background_gradient(cmap='YlOrBr',subset=["MR"])\
.background_gradient(cmap='bone_r',subset=["Positive"])
Out[71]:
In [72]:
df_test.corr().style.background_gradient(cmap='Blues').format("{:.2f}")
Out[72]:
In [73]:
fig = px.bar(df_test.sort_values("Cumulative total"),
x='country', y="Cumulative total",
text = "MR",
hover_name="country",
hover_data=["confirmed","deaths","Cumulative total","Positive"],
title='COVID-19: Tests Over Countries',
)
fig.update_xaxes(title_text="Country")
fig.update_yaxes(title_text="Number of Tests (Text on bars is MR %)")
fig.show()
In [74]:
fig = px.bar(df_test.sort_values("Positive"),
x='country', y="Positive",
text = "MR",
hover_name="country",
hover_data=["confirmed","deaths","Cumulative total","Positive"],
title='COVID-19: Test(Positive) Over Countries',
)
fig.update_xaxes(title_text="Country")
fig.update_yaxes(title_text="Positive Tests(%)- (Text on bars is MR %)")
fig.show()
In [75]:
fig = px.bar(df_test.sort_values("MR"),
x='country', y="MR",
text = "Positive",
hover_name="country",
hover_data=["confirmed","deaths","Cumulative total","Positive"],
title='COVID-19: MR Over Countries',
)
fig.update_xaxes(title_text="Country")
fig.update_yaxes(title_text="MR(%)- (Text on bars is Positive %)")
fig.show()
In [76]:
fig = px.bar(df_test.sort_values("Cumulative total per million"),
x='country', y="Cumulative total per million",
text = "MR",
hover_name="country",
hover_data=["confirmed","deaths","Cumulative total","Positive"],
title='COVID-19: Tests per million Over Countries',
)
fig.update_xaxes(title_text="Country")
fig.update_yaxes(title_text="Tests per million(%)- (Text on bars is MR %)")
fig.show()
In [77]:
fig = px.scatter(df_test, y=df_test.loc[:,"MR"],
x= df_test.loc[:,"Positive"],
color= "country", hover_name="country",
hover_data=["confirmed","deaths","Cumulative total"],
color_continuous_scale=px.colors.sequential.Plasma,
title='COVID-19: Test(Positive) vs Mortality rate',
size = np.power(df_test["confirmed"]+1,0.3)-0.5,
size_max = 30,
height =600,
)
fig.update_coloraxes(colorscale="hot")
fig.update(layout_coloraxis_showscale=False)
fig.update_yaxes(title_text="Mortality Rate (%)")
fig.update_xaxes(title_text="Tests Positive (%)")
fig.show()
Pie Viz¶
In [78]:
rows = 4
f = plt.figure(figsize=(15,10*rows))
stats = [df_covid19.loc[:,['country','Confirmed']],df_covid19.loc[:,['country','Deaths']],df_covid19.loc[:,['country','Active']],df_covid19.loc[:,['country','Recovered']]]
label = ["Confirmed","Deaths","Active","Recovered"]
threshold = [25000,1000,15000,5000]
for i, stat in enumerate(stats):
plt.tight_layout()
df_countries = stat.groupby(["country"]).sum()
df_countries = df_countries.sort_values(df_countries.columns[-1],ascending= False)
others = df_countries[df_countries[df_countries.columns[-1]] < threshold[i] ].sum()[-1]
df_countries = df_countries[df_countries[df_countries.columns[-1]] > threshold[i]]
df_countries = df_countries[df_countries.columns[-1]]
df_countries["others"] = others
labels = [df_countries.index[i] +" (" + str(int(df_countries[i])) +") "for i in range(df_countries.shape[0])]
ax = f.add_subplot(rows,1,i+1)
plt.pie(df_countries, labels=labels,autopct='%1.1f%%',pctdistance=0.85, labeldistance=1.1,textprops = {'fontsize':10.5})
my_circle=plt.Circle( (0,0), 0.7, color='white')
p=plt.gcf()
p.gca().add_artist(my_circle)
plt.text(0.5,0.5,"World Total "+label[i]+ " COVID-19 Cases\n"+str(stat.sum().values[1]), horizontalalignment='center',verticalalignment='center',transform=ax.transAxes, size=18, alpha = 0.6)
plt.show()
In [79]:
cols =1
rows = 2
f = plt.figure(figsize=(15,10*rows))
# SubPlot 1
ax = f.add_subplot(211)
plt.plot(np.sum(np.asarray(df_confirmed.iloc[:,5:]),axis = 0),np.sum(np.asarray(df_deaths.iloc[:,5:]),axis = 0))
axis_label = ["Cumulative Confirmed Cases","Cumulative Deaths"]
plt_title = "COVID-19: World - \nCumulative Confirmed Cases Vs Cumulative Deaths Curve"
plot_params(ax,axis_label,plt_title)
# # SubPlot 2
# ax = f.add_subplot(212)
# mortality_rate = get_mortality_rate(df_confirmed,df_deaths,continent=continents[0])
# plt.plot(np.sum(np.asarray(df_recovered.iloc[:,5:]),axis = 0),np.sum(np.asarray(df_deaths.iloc[:,5:]),axis = 0))
# axis_label = ["Cumulative Recoveries","Cumulative Deaths"]
# plt_title = "COVID-19: World - Cumulative Recovery Vs Cumulative Deaths Curve"
# plot_params(ax,axis_label,plt_title)
plt.minorticks_on()
plt.savefig(out+'Cumulative Confirmed Cases Vs Cumulative Deaths Curve.png')
plt.show()
USA¶
In [81]:
date_usa = datetime.strptime(df_confirmed.columns[-1],'%m/%d/%y').strftime("%m-%d-%Y")
df_temp = pd.read_csv("https://raw.githubusercontent.com/CSSEGISandData/COVID-19/master/csse_covid_19_data/csse_covid_19_daily_reports/"+date_usa+".csv")
df_usa = df_temp.loc[df_temp["Country_Region"]== "US"]
df_usa = df_usa.rename(columns={"Admin2":"County"})
In [82]:
total = df_usa.sum()
total.name = "Total"
pd.DataFrame(total).transpose().loc[:,["Confirmed","Deaths"]].style.background_gradient(cmap='Purples',axis=1)
Out[82]:
In [83]:
df_usa.loc[:,["Confirmed","Deaths","Province_State"]].groupby(["Province_State"]).sum().sort_values("Confirmed",ascending=False).style.background_gradient(cmap='Blues',subset=["Confirmed"]).background_gradient(cmap='Reds',subset=["Deaths"])
Out[83]:
In [84]:
df_usa.loc[:,["Province_State","Confirmed","Deaths","County"]].sort_values("Confirmed",ascending=False).set_index("County")[:10].style.background_gradient(cmap='Blues',subset=["Confirmed"])\
.background_gradient(cmap='Reds',subset=["Deaths"])
Out[84]:
In [85]:
f = plt.figure(figsize=(10,5))
f.add_subplot(111)
plt.axes(axisbelow=True)
plt.barh(df_usa.groupby(["Province_State"]).sum().sort_values('Confirmed')["Confirmed"].index[-10:],df_usa.groupby(["Province_State"]).sum().sort_values('Confirmed')["Confirmed"].values[-10:],color="darkcyan")
plt.tick_params(size=5,labelsize = 13)
plt.xlabel("Confirmed Cases",fontsize=18)
plt.title("Top 10 States: USA (Confirmed Cases)",fontsize=20)
plt.grid(alpha=0.3)
plt.savefig(out+'Top 10 States_USA (Confirmed Cases).png')
In [86]:
f = plt.figure(figsize=(10,5))
f.add_subplot(111)
plt.axes(axisbelow=True)
plt.barh(df_usa.groupby(["Province_State"]).sum().sort_values('Deaths')["Deaths"].index[-10:],df_usa.groupby(["Province_State"]).sum().sort_values('Deaths')["Deaths"].values[-10:],color="crimson")
plt.tick_params(size=5,labelsize = 13)
plt.xlabel("Deaths",fontsize=18)
plt.title("Top 10 States: USA (Deaths Cases)",fontsize=20)
plt.grid(alpha=0.3)
plt.savefig(out+'Top 10 States_USA (Deaths Cases).png')
In [87]:
df_usa = df_usa.replace(np.nan, 0, regex=True)
usa = folium.Map(location=[37, -102], zoom_start=4,max_zoom=9,min_zoom=4)
for i in np.int32(np.asarray(df_usa[df_usa['Confirmed'] > 0].index)):
folium.Circle(
location=[df_usa.loc[i]['Lat'], df_usa.loc[i]['Long_']],
tooltip = "<h5 style='text-align:center;font-weight: bold'>"+df_usa.loc[i]['Province_State']+"</h5>"+
"<div style='text-align:center;'>"+str(np.nan_to_num(df_usa.loc[i]['County']))+"</div>"+
"<hr style='margin:10px;'>"+
"<ul style='color: #444;list-style-type:circle;align-item:left;padding-left:20px;padding-right:20px'>"+
"<li>Confirmed: "+str(df_usa.loc[i]['Confirmed'])+"</li>"+
"<li>Active: "+str(df_usa.loc[i]['Active'])+"</li>"+
"<li>Recovered: "+str(df_usa.loc[i]['Recovered'])+"</li>"+
"<li>Deaths: "+str(df_usa.loc[i]['Deaths'])+"</li>"+
"<li>Mortality Rate: "+str(np.round(df_usa.loc[i]['Deaths']/(df_usa.loc[i]['Confirmed']+1)*100,2))+"</li>"+
"</ul>"
,
radius=int((np.log2(df_usa.loc[i]['Confirmed']+1))*6000),
color='#ff6600',
fill_color='#ff8533',
fill=True).add_to(usa)
usa
Out[87]:
In [88]:
state_geo = requests.get('https://raw.githubusercontent.com/python-visualization/folium/master/examples/data/us-states.json').json()
county_geo = requests.get('https://raw.githubusercontent.com/python-visualization/folium/master/examples/data/us_counties_20m_topo.json').json()
# county_geo
In [89]:
data_temp = df_usa.groupby(["Province_State"]).sum().reset_index().drop(["Lat","Long_"],axis=1)
data_temp["Confirmed_log"] = np.log10(data_temp["Confirmed"]+1)
bins = list(data_temp['Confirmed_log'].quantile([0, 0.25, 0.5, 0.75,0.95 ,1]))
m = folium.Map(location=[37, -102], zoom_start=4,max_zoom=6,min_zoom=3)
# Add the color for the chloropleth:
folium.Choropleth(
geo_data=state_geo,
name='choropleth',
data = data_temp,
columns=['Province_State', 'Confirmed_log'],
key_on='feature.properties.name',
fill_color='Reds',
fill_opacity=0.7,
line_opacity=0.2,
bins = bins,
reset=True,
legend_name='Confirmed cases log (10^x)'
).add_to(m)
folium.LayerControl().add_to(m)
legend_html = "<div style='padding:10px;background-color:rgba(255,255,255,0.5);position:fixed;bottom:20px;left:20px;z-index:1000'>"
legend_html += "<div style='width:100%;text-align:center;'><h4>Index for Legend</h4></div><hr style='border-top-color: rgba(25,25,25,0.5);'>"
legend_html += "<ul style='margin:0;padding:0;color: #555;list-style-type:circle;align-item:left;padding-left:20px;padding-right:20px'>"
for i in bins:
legend_html += "<li style='margin:0;padding:0;line-height: 0;'>Value "+str(np.round(i,2))+" : "+str(int(10**i)-1)+"</li><br>"
legend_html += "</ul></div>"
m.get_root().html.add_child(folium.Element(legend_html))
m
Out[89]:
In [90]:
data_temp = df_usa.groupby(["FIPS"]).sum().reset_index().drop(["Lat","Long_"],axis=1)
data_temp["Confirmed_log"] = np.log10(data_temp["Confirmed"]+1)
df_usa_series = data_temp.set_index('FIPS')['Confirmed_log']
colorscale = branca.colormap.linear.Reds_09.scale(0,data_temp["Confirmed_log"].max()-1)
# print(df_usa_series.max())
def style_function(feature):
employed = df_usa_series.get(int(feature['id'][-5:]), 0)
return {
'fillOpacity': 0.5,
'weight': 0,
'fillColor': '#black' if employed is None else colorscale(employed)
}
m = folium.Map(
location=[37, -102],
tiles='cartodbpositron',
zoom_start=4,
min_zoom=3,
max_zoom=7
)
folium.TopoJson(
county_geo,
'objects.us_counties_20m',
style_function=style_function
).add_to(m)
m
Out[90]:
In [91]:
# Visible = Input(shape=(1,))
# Dense_l1 = Dense(80,name="Dense_l1")(Visible)
# LRelu_l1 = LeakyReLU(name = "LRelu_l1")(Dense_l1)
# Dense_l2 = Dense(80,name = "Dense_l2")(LRelu_l1)
# LRelu_l2 = LeakyReLU(name = "LRelu_l2")(Dense_l2)
# Dense_l3 = Dense(1,name="Dense_l3")(LRelu_l2)
# LRelu_l3 = LeakyReLU(name = "Output")(Dense_l3)
# model_usa_c = models.Model(inputs=Visible, outputs=LRelu_l3)
# model_usa_c.compile(optimizer=Adam(lr=0.0001),
# loss='mean_squared_error',
# metrics=['accuracy'])
# model_usa_c.summary()
In [92]:
df_temp = df_confirmed.groupby(["country"]).sum()
df_temp = np.asarray(df_temp[df_temp.index == "USA"].iloc[:,2:])[0]
data_y = np.log10(df_temp).astype("float32")
data_x = np.arange(1,len(data_y)+1)
# data_x.shape[0]
In [93]:
model_usa_c = models.load_model("model_usa_c_v2.h5")
model_usa_c.summary()
In [94]:
M = 1000000
prediction_days = 7
df_temp = df_confirmed.groupby(["country"]).sum()
temp_data = np.asarray(df_temp[df_temp.index == "USA"].iloc[:,2:])[0]
data = np.power(10,model_usa_c.predict(np.arange(1,len(temp_data)+prediction_days+1)))
# temp_data = df_confirmed.iloc[:,5:].sum(axis =0)
f = plt.figure(figsize=(12,8))
ax = f.add_subplot(111)
date = np.arange(0,len(temp_data))
marker_style = dict(linewidth=3, linestyle='-', marker='o',markersize=7, markerfacecolor='#ffffff')
plt.plot(date,temp_data,"-.",color="darkcyan",**marker_style, label="Actual Curve")
date = np.arange(0,len(data))
plt.plot(date,data,"-.",color="orangered",label="Predicted Curve")
nextdays = [(datetime.strptime(d[-1],'%d %b')+timedelta(days=i)).strftime("%d %b") for i in range(1,prediction_days+1)]
total =d+nextdays
text = "Prediction for next "+str(prediction_days) +" days:\n"
for i in range(prediction_days):
text += nextdays[i]+" : "+str(np.round(data[-1*(prediction_days-i)],-3)[0]/M)+" M\n"
plt.text(0.02, 0.78, text, fontsize=17, horizontalalignment='left', verticalalignment='top', transform=ax.transAxes,bbox=dict(facecolor='white', alpha=0.4))
# X-axis
plt.xticks(list(np.arange(0,len(total),int(len(total)/5))),total[:-1:int(len(total)/5)]+[total[-1]])
# Tick-Parameters
ax.xaxis.set_minor_locator(ticker.AutoMinorLocator())
ax.yaxis.set_minor_locator(ticker.AutoMinorLocator())
ax.tick_params(which='both', width=1,labelsize=14)
ax.tick_params(which='major', length=6)
ax.tick_params(which='minor', length=3, color='0.8')
# Grid
plt.grid(lw = 1, ls = '-', c = "0.7", which = 'major')
plt.grid(lw = 1, ls = '-', c = "0.9", which = 'minor')
# Plot Title
plt.title("COVID-19 Next 7 day Prediction Curve-USA Confirmed Cases",{'fontsize':22})
# Axis Lable
plt.xlabel("Date",fontsize =18)
plt.ylabel("Number of Confirmed Cases",fontsize =18)
plt.legend(fontsize =18)
plt.tick_params(labelsize = 13)
plt.savefig(out+"Prediction Curve-Confirmed (USA).png")
plt.show()
In [ ]: