# Current time: 
TODAY <- today()

# Data links (source: https://github.com/CSSEGISandData/COVID-19): 
link1 <- "https://raw.githubusercontent.com/CSSEGISandData/COVID-19/master/csse_covid_19_data/csse_covid_19_time_series/time_series_19-covid-Recovered.csv"


## Confrimed 
link2 <- "https://raw.githubusercontent.com/CSSEGISandData/COVID-19/master/csse_covid_19_data/csse_covid_19_time_series/time_series_19-covid-Confirmed.csv"


## DEATh 
link3 <- "https://raw.githubusercontent.com/CSSEGISandData/COVID-19/master/csse_covid_19_data/csse_covid_19_time_series/time_series_19-covid-Deaths.csv"

link2 %>% 
  read_csv() %>% 
  rename(province = `Province/State`, country = `Country/Region`) %>% 
  gather(Date, Confirmed, -province, - country, - Lat, -Long) %>% 
  mutate(Date = mdy(Date)) -> df_confirmed

link3 %>% 
  read_csv() %>% 
  rename(province = `Province/State`, country = `Country/Region`) %>% 
  gather(Date, Confirmed, -province, - country, - Lat, -Long) %>% 
  mutate(Date = mdy(Date)) -> df_death

df_confirmed %>% 
  group_by(country, Date) %>% 
  summarise(confirmed = sum(Confirmed)) %>% 
  ungroup() -> Con



df_death %>% 
  group_by(country, Date) %>% 
  summarise(death = sum(Confirmed)) %>% 
  ungroup() ->Death


full_join(Con, Death, by =c("Date" ,"country")) -> ConDeath

ConDeath %>% 
  filter(country == "Italy") -> italy

# Train Test 나눠주기 

train <- italy$death[1:50] %>% ts()
test <- italy$death[51:56] %>% ts()

h<- length(test)

my_arima <- auto.arima(train)

# Model 

predicted_arima <- forecast(my_arima, h=h)$mean %>%  as.vector()

tibble(Date = italy$Date[51:56],
           Actual = test %>%  as.vector(),
           Predicted = predicted_arima %>% round(0),
           Error = Predicted - Actual,
           Error_Percent = round(Error/ Actual,2)) -> df_arima

datatable(df_arima)

train_full <- italy$death %>%  ts()
my_arima_full <- auto.arima(train_full)

# Use the model for forecasting for next 20 days from 2020-03-17: 
predicted_arima <- forecast(my_arima_full, h=20)$mean %>%  as.vector()


arima_next20 <- data_frame(Date = italy$Date %>%  max() + 1:20,
                           Predicted = predicted_arima %>%  round(0)) 


arima_next20 %>% 
  knitr::kable(caption = "Table:Deaths by Coronavirus Predicted for Next 20 Days ")

arima_next20 %>% 
  rename(death = Predicted) %>% 
  mutate(Type ="Predicted") -> predict_table



italy %>% 
  mutate(Type ="Actual") -> actual_table

  

ggplot()+
  geom_point(data=predict_table, aes(Date, death, color= Type), size=2 )+
  geom_point(data=actual_table, aes(Date, death, color= Type ), size=2)+
  labs(title = "Death by Corona Predicted for next 20 days in Italy")+
  scale_color_manual(values = c("cyan", "red")) +
  theme_modern_rc()+
  theme(axis.text.y = element_text(color = "white", size = 12)) + 
  theme(axis.text.x = element_text(color = "white", size = 12)) + 
  theme(axis.title.x = element_blank()) + 
  theme(axis.title.y = element_blank()) -> plot_italy
    

plotly:: ggplotly(plot_italy)

ConDeath %>% 
  filter(country == "Korea, South") -> KOREA

# train test 나누기 
train <- KOREA$confirmed[1:50] %>% ts()
test <- KOREA$confirmed[51:56] %>% ts()

# h 정해주기 
h <- length(test)

# model auto <- train 으로 학습 
korea_arima <- auto.arima(train)

# 예측 모델링 
predict_korea_confrim <- forecast(korea_arima, h=h)$mean %>%  as.vector()

# 데이터 프레임화 시켜주기 
tibble(Date = KOREA$Date[51:56],
       Actual = test %>% as.vector(),
       Predict = predict_korea_confrim %>%  round(0),
       Error = Predict- Actual,
       Error_Percent = round(Error/ Actual,2)) -> df_Korea

knitr::kable(df_Korea)

train_all <- KOREA$confirmed %>%  ts()
Arima_full <- auto.arima(train_all)


pred_arima <- forecast(Arima_full, h=20)$mean %>% as.vector()

next_20_Korea_confirm <- data_frame(Date= KOREA$Date %>% max()+1:20,
                                    Predict = pred_arima %>%  round(0))

knitr::kable(next_20_Korea_confirm)

next_20_Korea_confirm %>% 
  rename(Confirmed = Predict) %>% 
  mutate(Type = "Predicted") -> KOREA_predict



KOREA %>% 
  mutate(Type ="Actual") -> KOREA_actual

KOREA_actual

ggplot()+
  geom_point(data= KOREA_predict, aes(Date,Confirmed, color=Type ), size=2)+
  geom_point(data= KOREA_actual, aes(Date, confirmed, color=Type), size=2)+
  theme_modern_rc()+
  scale_y_continuous(limits = c(0,10000), 
                     breaks = seq(0,10000,1000))+
  scale_color_manual(values = c("cyan", "red")) + 
  labs(title ="Predicting confrimed cases for next 20 days in South Korea")+
  theme(plot.title = element_text(size = 15)) -> KOREA_plot

plotly::ggplotly(KOREA_plot)