-
Notifications
You must be signed in to change notification settings - Fork 0
/
flu_forecast.R
49 lines (40 loc) · 1.29 KB
/
flu_forecast.R
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
library(tidyverse)
library(prophet)
library(Hmisc)
df <- fluviewr_data()
df <- subset(df, df$week%nin%seq(19,39))
df %>%
mutate(
pct_flu = total_flu/total_specimens*100
) %>%
select(c(wk_date,pct_flu)) %>%
rename(
ds = "wk_date",
y = "pct_flu"
) -> df
### split before/after COVID
df.pre <- subset(df, df$ds<"2020-09-27")
df.post<- data.frame(cbind(
"ds" = c(rep(NA, times=nrow(df.pre)), df$ds[df$ds>="2020-09-27"]),
"y" = c(rep(NA, times=nrow(df.pre)), df$y[df$ds>="2020-09-27"])))
df.post$ds <- df$ds
####################################################################
####################################################################
####################################################################
### set up model
m <- prophet(daily.seasonality= F,
weekly.seasonality = T,
yearly.seasonality = T,
interval.width = .8)
## add us holidays
m <- add_country_holidays(m, country_name = 'US')
## fit
m <- fit.prophet(m, df.pre)
# forecast
future <- make_future_dataframe(m, periods = 1000)
forecast <- predict(m, future)
plot(m, forecast)
p<-plot(m, forecast)
t <- p +
geom_point(data=df.post, aes(x=as.POSIXct(df.post$ds), y=y, colour="actual"))
plotly::ggplotly(t)