By week

library(readxl)
library(DescTools)
require(forecast)
library(xlsx)
library('zoo')
require(dplyr)
require(broom)
HHC<-read_excel('C:/Users/Tung.Nguyen/Documents/Work_Unilever/Transformation project/Machine Learning for Forecast/Raw new/Old/HHC test - by week.xlsx', sheet = "Raw_Actual")
head(HHC)
#create df forecast result to store data later on
forecast_result <-data.frame( "forecast"=c(0) , "SKU"=c(0), "method" = c(0))
forecast_result

#Input horizon
horizon <- 104



#Create function for seasonal linear
Forecast.SL <- function(ts_raw, horizon){
    #fit model
    FC_SL <- tslm(ts_raw ~ season + trend)
    FC.Mod <- forecast(FC_SL, h=horizon)
    result <- as.data.frame(FC.Mod)
    plot(FC.Mod, main = colnames(raw[,2]))
    dev.new()
    summary(FC.Mod)
    #create data frame result
    xxx <- data.frame("date"=result[,0],"forecast" = result[,1], "SKU" = colnames(raw[,2]), "method" = "SLN")
    return (xxx)
}

#Create function for HoltWinter
Forecast.HW <- function(ts_raw, horizon){
    #fit model
    FC_HW <- HoltWinters(ts_raw)
    FC.Mod <- forecast(FC_HW, h=horizon)
    result <- as.data.frame(FC.Mod)
    plot(FC.Mod, main = colnames(raw[,2]))

    #create data frame result
    xxx <- data.frame("date"=result[,0],"forecast" = result[,1], "SKU" = colnames(raw[,2]), "method" = "HW")
        
    return (xxx)
    #print(forecast_result)
}



#Go through each SKU
for (i in 2:ncol(HHC)){
raw <- HHC[,c(1,i)]

    raw2 = as.vector(t(raw[,2]))
    ts_r <- ts(raw2, start = c(2016,7,1), frequency = 52)
    ts_r <- tsclean(ts_r,replace.missing = TRUE, lambda = NULL)
    plot(ts_r)
    final <- Forecast.SL(ts_r, horizon)

forecast_result<-rbind(forecast_result,final)
print("done 1 SKU")
}

#write.xlsx(forecast_result,file = 'C:/Users/Tung.Nguyen/Documents/Work_Unilever/Transformation project/Machine Learning for Forecast/result_byweek_Omo.xlsx')
#print("successfull")