# This R environment comes with all of CRAN preinstalled, as well as many other helpful packages
# The environment is defined by the kaggle/rstats docker image: https://github.com/kaggle/docker-rstats
# For example, here's several helpful packages to load in 

library(ggplot2) # Data visualization
library(readr) # CSV file I/O, e.g. the read_csv function

# Input data files are available in the "../input/" directory.
# For example, running this (by clicking run or pressing Shift+Enter) will list the files in the input directory

system("ls ../input")

# Any results you write to the current directory are saved as output.
 library(data.table)
library(Metrics)
library(dplyr)
library(knitr)
n = 15e6
 
library(data.table)

print("Reading data")
train <- fread('../input/train.csv', 
               select = c('Cliente_ID', 'Producto_ID', 'Agencia_ID', 'Ruta_SAK', 'Demanda_uni_equil'))

test <- fread('../input/test.csv', 
              select = c('id', 'Cliente_ID', 'Producto_ID', 'Agencia_ID', 'Ruta_SAK'))

print("Computing means")
#transform target variable to log(1 + demand) - this makes sense since we're 
#trying to minimize rmsle and the mean minimizes rmse:
train$log_demand = log1p(train$Demanda_uni_equil)
train$log_demand
mean_total <- mean(train$log_demand) #overall mean
#mean by product
mean_P <-  train[, .(MP = mean(log_demand)), by = .(Producto_ID)]
#mean by product and ruta
mean_PR <- train[, .(MPR = mean(log_demand)), by = .(Producto_ID, Ruta_SAK)] 
#mean by product, client, agencia
mean_PCA <- train[, .(MPCA = mean(log_demand)), by = .(Producto_ID, Cliente_ID, Agencia_ID)]

print("Merging means with test set")

submit<-merge(test,mean_PCA,all.x=TRUE,by=c("Producto_ID", "Cliente_ID", "Agencia_ID"))
submit <- merge(submit, mean_PR, all.x = TRUE, by = c("Producto_ID", "Ruta_SAK"))
submit <- merge(submit, mean_P, all.x = TRUE, by = "Producto_ID")

submit