list_packages <- c("forecast", "readxl", "stargazer", "fpp",
"fpp2", "scales", "quantmod", "urca",
"vars", "tseries", "ggplot2", "dplyr", 'tidyverse', 'lubridate', 'scales')
new_packages <- list_packages[!(list_packages %in% installed.packages()[,"Package"])]
if(length(new_packages)) install.packages(new_packages)
# Load necessary packages
lapply(list_packages, require, character.only = TRUE)## [[1]]
## [1] TRUE
##
## [[2]]
## [1] TRUE
##
## [[3]]
## [1] TRUE
##
## [[4]]
## [1] TRUE
##
## [[5]]
## [1] TRUE
##
## [[6]]
## [1] TRUE
##
## [[7]]
## [1] TRUE
##
## [[8]]
## [1] TRUE
##
## [[9]]
## [1] TRUE
##
## [[10]]
## [1] TRUE
##
## [[11]]
## [1] TRUE
##
## [[12]]
## [1] TRUE
##
## [[13]]
## [1] TRUE
##
## [[14]]
## [1] TRUE
##
## [[15]]
## [1] TRUEsales <- read_excel("/Users/saurabh/Downloads/Time Series /Final Project/global_superstore_2016.xlsx", sheet = 'Orders')head(sales)## # A tibble: 6 x 24
## `Row ID` `Order ID` `Order Date` `Ship Date` `Ship Mode`
## <dbl> <chr> <dttm> <dttm> <chr>
## 1 40098 CA-2014-A… 2014-11-11 00:00:00 2014-11-13 00:00:00 First Class
## 2 26341 IN-2014-J… 2014-02-05 00:00:00 2014-02-07 00:00:00 Second Cla…
## 3 25330 IN-2014-C… 2014-10-17 00:00:00 2014-10-18 00:00:00 First Class
## 4 13524 ES-2014-K… 2014-01-28 00:00:00 2014-01-30 00:00:00 First Class
## 5 47221 SG-2014-R… 2014-11-05 00:00:00 2014-11-06 00:00:00 Same Day
## 6 22732 IN-2014-J… 2014-06-28 00:00:00 2014-07-01 00:00:00 Second Cla…
## # … with 19 more variables: `Customer ID` <chr>, `Customer Name` <chr>,
## # Segment <chr>, `Postal Code` <dbl>, City <chr>, State <chr>,
## # Country <chr>, Region <chr>, Market <chr>, `Product ID` <chr>,
## # Category <chr>, `Sub-Category` <chr>, `Product Name` <chr>,
## # Sales <dbl>, Quantity <dbl>, Discount <dbl>, Profit <dbl>, `Shipping
## # Cost` <dbl>, `Order Priority` <chr>Wrangling ->Dates ->Aggregate to Country - Store - Month level Country - Product - Month level Country - Month level ->Drop unnecessary columns keep only one size
EDA -> Time series pattern by countries -> Heat map by stores; sizes
cat('Number of rows:', nrow(sales), '\n')## Number of rows: 51290cat('Columns names are :- \n')## Columns names are :-#sales <- sales[!duplicated(sales),]
names(sales)## [1] "Row ID" "Order ID" "Order Date" "Ship Date"
## [5] "Ship Mode" "Customer ID" "Customer Name" "Segment"
## [9] "Postal Code" "City" "State" "Country"
## [13] "Region" "Market" "Product ID" "Category"
## [17] "Sub-Category" "Product Name" "Sales" "Quantity"
## [21] "Discount" "Profit" "Shipping Cost" "Order Priority"sapply(sales, function(x) sum(is.na(x)))## Row ID Order ID Order Date Ship Date Ship Mode
## 0 0 0 0 0
## Customer ID Customer Name Segment Postal Code City
## 0 0 0 41296 0
## State Country Region Market Product ID
## 0 0 0 0 0
## Category Sub-Category Product Name Sales Quantity
## 0 0 0 0 0
## Discount Profit Shipping Cost Order Priority
## 0 0 0 0sales <- subset(sales, select = c( `Order Date`, Country, Market, Category, `Sub-Category`, `Product Name`, `Product ID`, Quantity))
names(sales)## [1] "Order Date" "Country" "Market" "Category"
## [5] "Sub-Category" "Product Name" "Product ID" "Quantity"sales$Year_mon <- format(sales$`Order Date`, '%Y-%m')
sales$Year <- format(sales$`Order Date`, '%Y')
sales$Month <- format(sales$`Order Date`, '%B')
sales$Date <- lubridate::floor_date(sales$`Order Date`, unit = 'month')Country_filtered <- sales[which(sales$Country %in% c('United States', 'Australia', 'France', 'Mexico', 'Germany', 'China')),]picked_country <- 'United States'
Country_picked <- sales[which(Country_filtered$Country %in% c(picked_country)),]# Summing up transactions to a Country - shop - product - gender - size - year - month level
Country_level <- aggregate(cbind( Quantity) ~ Date, data = Country_picked, FUN = sum)
Category_sales <- aggregate(cbind(Quantity) ~ Category + Date, data = Country_picked, FUN = sum)
Sub_Category_sales <- aggregate(cbind(Quantity) ~ Category + `Sub-Category` + Date, data = Country_picked, FUN = sum)Country_filtered <- sales[which(sales$Country %in% c('United States', 'Australia', 'France', 'Mexico', 'Germany', 'China')),]
Country_picked <- sales[which(Country_filtered$Country %in% c(picked_country)),]
Country_level <- aggregate(cbind( Quantity) ~ Date, data = Country_picked, FUN = sum)
Category_sales <- aggregate(cbind(Quantity) ~ Category + Date, data = Country_picked, FUN = sum)
Sub_Category_sales <- aggregate(cbind(Quantity) ~ Category + `Sub-Category` + Date, data = Country_picked, FUN = sum)EDA Phase
1. Monthly time series plot
ggplot(Country_level, aes(y = Quantity, x = Date)) +
scale_x_datetime(labels = date_format("%b"), date_breaks = '1 month', expand = c(0.05,0.1)) +
facet_grid(~year(Date), space = "free_x", scales = "free_x", switch='x')+
geom_point() + geom_line(linetype = 'dashed') +
ggtitle("Monthly Time Series Plot") +
theme_bw() +
theme(
plot.title = element_text(hjust = 0.5),
strip.placement = 'outside',
strip.background = element_rect( colour = "grey50"),
#panel.spacing = unit(0,'cm'),
axis.text.x = element_text(angle = 90)
) 
Observations from the plot: - * There is a discernable time series pattern. * There is obvious monthly seasonality. * There is a trend component.
2. Monthly Series for each category
ggplot(Category_sales, aes(y = Quantity, x = Date)) +
scale_x_datetime(labels = date_format("%b"), date_breaks = '1 month', expand = c(0.05,0.1)) +
facet_grid(~year(Date), space = "free_x", scales = "free_x", switch='x')+
geom_point() + geom_line( aes(color = Category,linetype = Category), size = 1) +
ggtitle("Monthly Time Series Plot by Category") +
theme_bw() +
theme(
plot.title = element_text(hjust = 0.5),
strip.placement = 'outside',
strip.background = element_rect( colour = "grey50"),
#panel.spacing = unit(0,'cm'),
axis.text.x = element_text(angle = 90)
) 
Observations from the plot: - * Office supplies bought are considerably higher in volume than Furniture and Technology. * Office supplies also begin at a similar point in the beginning of the year but rise in volume much faster. * Furniture and technology volumes and patterns are very close to each other.
3. Sub Category heat map
ggplot(Sub_Category_sales, aes(y = `Sub-Category`, x = Date, fill = Quantity)) +
scale_x_datetime(labels = date_format("%b"), date_breaks = '1 month', expand = c(0,0)) +
facet_grid(~year(Date), space = "free_x", scales = "free_x", switch='x') +
geom_tile() +
geom_text(aes(label = Quantity), size = 3) +
scale_fill_gradient(low="grey", high="red") +
ggtitle("Monthly Time Series Heat Map by Sub-Category") +
theme_bw() +
theme(
plot.title = element_text(hjust = 0.5),
strip.placement = 'outside',
strip.background = element_rect( colour = "grey50"),
#panel.spacing = unit(0,'cm'),
axis.text.x = element_text(angle = 90)
) 
Observations from the plot: - * Most sub-categories have the same pattern * Binders and papers seem to be bought in the highest volume.
Forecast Model Identification
Sales_ts <- ts(Country_level$Quantity, frequency = 12, start = c(2012,1), end = c(2015,12))
Sales_ts## Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
## 2012 363 397 400 411 489 663 380 734 814 729 720 982
## 2013 404 299 524 476 645 993 514 989 869 820 1151 1080
## 2014 522 474 607 581 962 1329 783 1044 1367 814 1349 1316
## 2015 637 643 863 788 898 1300 668 1362 1572 1310 1836 1606Sales_train_ts <- ts(Country_level$Quantity[1:36], frequency = 12, start = c(2012,1), end = c(2014,12))
Sales_train_ts## Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
## 2012 363 397 400 411 489 663 380 734 814 729 720 982
## 2013 404 299 524 476 645 993 514 989 869 820 1151 1080
## 2014 522 474 607 581 962 1329 783 1044 1367 814 1349 1316Sales_train_ts %>% ggtsdisplay(lag.max =36)
Observations from the plot :-
- Spike at lag 12, there is correlation with the 12th lag element each time, indicating seasonality.
- The ACF at lags 12, 24 and 36 are decreasing slowly, so we can take a seasonal 1st order differencing to make it stationary
adf.test(Sales_train_ts, k =1)##
## Augmented Dickey-Fuller Test
##
## data: Sales_train_ts
## Dickey-Fuller = -3.488, Lag order = 1, p-value = 0.06047
## alternative hypothesis: stationaryadf.test(Sales_train_ts, k = 2)##
## Augmented Dickey-Fuller Test
##
## data: Sales_train_ts
## Dickey-Fuller = -2.3414, Lag order = 2, p-value = 0.4403
## alternative hypothesis: stationaryadf.test(Sales_train_ts, k = 12)##
## Augmented Dickey-Fuller Test
##
## data: Sales_train_ts
## Dickey-Fuller = -2.354, Lag order = 12, p-value = 0.4354
## alternative hypothesis: stationaryObservations * We see that since p-value is > 0.05, we fail to reject the null hypothesis and thus the series is non-stationary
Sales_train_ts %>%diff() %>% ggtsdisplay(lag.max =40)
Observations from the plot :-
- Spike at lag 12, there is correlation with the 12th lag element each time, indicating seasonality.
- The ACF at lags 12, 24 and 36 are decreasing slowly, so we can take a seasonal 1st order differencing to make it stationary
# Sales_train_ts %>% diff() %>% diff(lag=12) %>% ggtsdisplay(lag.max=36)
Sales_train_ts %>%diff() %>% diff(lag=12) %>% ggtsdisplay(lag.max=36)
Observations: - * The spike at lag 1 suggests an MR(1) component and the spike at lag 12 suggests a seasonal MR(1) component.
Sales_train_ts %>% log() %>% diff() %>% diff(lag=12) %>% adf.test()##
## Augmented Dickey-Fuller Test
##
## data: .
## Dickey-Fuller = -3.5363, Lag order = 2, p-value = 0.05885
## alternative hypothesis: stationaryIt is now stationary as the p-value is < 0.05 So when we perform seasonal difference and normal difference, the series is stationary.
Result of model identification
From the exploration, our guess is that the right model to fit this data is an ARIMA(0,1,1)(0,1,1)[12]
Let us check what auto arima tells us: -
auto_fit <- Sales_train_ts %>% auto.arima()## Warning: The chosen seasonal unit root test encountered an error when testing for the second difference.
## From stl(): series is not periodic or has less than two periods
## 1 seasonal differences will be used. Consider using a different unit root test.summary(auto_fit)## Series: .
## ARIMA(0,0,0)(0,1,0)[12] with drift
##
## Coefficients:
## drift
## 14.1181
## s.e. 2.3527
##
## sigma^2 estimated as 19961: log likelihood=-152.36
## AIC=308.72 AICc=309.3 BIC=311.08
##
## Training set error measures:
## ME RMSE MAE MPE MAPE MASE
## Training set 0.166133 112.9286 75.21243 -3.785049 10.53844 0.422344
## ACF1
## Training set -0.1480112The difference is that the auto arima model is missing the MR(1) seasonal component
Model Fitting
fit <- Sales_train_ts %>% Arima( order = c(1,1,2), seasonal = c(0,1,1), include.constant = FALSE, lambda = 0)
summary(fit)## Series: .
## ARIMA(1,1,2)(0,1,1)[12]
## Box Cox transformation: lambda= 0
##
## Coefficients:
## ar1 ma1 ma2 sma1
## -0.9997 0.0800 -0.8794 -0.5829
## s.e. 0.0031 0.3845 0.3795 0.6176
##
## sigma^2 estimated as 0.02771: log likelihood=7.12
## AIC=-4.23 AICc=-0.7 BIC=1.45
##
## Training set error measures:
## ME RMSE MAE MPE MAPE MASE
## Training set 22.37088 100.3341 65.73635 2.151166 8.020521 0.3691325
## ACF1
## Training set 0.0008944884Checking terms for significance
-0.6235/0.1787 ## [1] -3.4890880.2691/0.3991## [1] 0.6742671sma1 comes out as not significant as the absolute value of the t-statistic is < 1.96 This is also what the auto arima model suggested to us. Let’s then compare the AICs of the two models.
AIC(fit)## [1] -4.231914AIC(auto_fit)## [1] 308.7243The (0,1,1)(0,1,0)[12] which the auto arima suggested has a slightly lower AIC
summary(auto_fit)## Series: .
## ARIMA(0,0,0)(0,1,0)[12] with drift
##
## Coefficients:
## drift
## 14.1181
## s.e. 2.3527
##
## sigma^2 estimated as 19961: log likelihood=-152.36
## AIC=308.72 AICc=309.3 BIC=311.08
##
## Training set error measures:
## ME RMSE MAE MPE MAPE MASE
## Training set 0.166133 112.9286 75.21243 -3.785049 10.53844 0.422344
## ACF1
## Training set -0.1480112Checking significance
-0.6097/0.1636## [1] -3.726773The term is significant as the absolute value of the test statistic is > 1.96
Residual Diagnostics and Model Adequacy
checkresiduals(fit)
##
## Ljung-Box test
##
## data: Residuals from ARIMA(1,1,2)(0,1,1)[12]
## Q* = 5.2466, df = 3, p-value = 0.1546
##
## Model df: 4. Total lags used: 7The residuals look good. There are no spikes in the ACF, there are no obvious patterns in the residuals and the residuals are centered around 0.
In the Ljung-Box test, we see that the p-value is > 0.05 which means we cannot reject the null hypothesis. Thus the model is adequate.
Forecasting next 12 months
forecasts <- fit %>% forecast(h = 12)
# MSE and MAE
cat('MSE of the model is:', mean((forecasts$mean[1:12] - Country_level$Quantity[37:48])^2), '\n')## MSE of the model is: 20956.52cat('MAE of the model is:', sum(abs(forecasts$mean[1:12] - Country_level$Quantity[37:48])), '\n')## MAE of the model is: 1443.797forecasts <- fit %>% forecast(h = 12)
forecast_values <- vector()
forecast_values <- c(forecast_values, rep(NA, 48))
forecast_values[37:48] <- forecasts$mean[1:12]
plotting_dataset <- Country_level
plotting_dataset$Forecasts <- forecast_values
names(plotting_dataset)[2] <- 'Actuals'
ggplot(plotting_dataset, aes(x = Date)) +
geom_line(aes(y = Actuals)) +
geom_line(aes(y = Forecasts), color = 'Red', linetype = 'dashed', size = 1)## Warning: Removed 36 rows containing missing values (geom_path).
forecasts <- auto_fit %>% forecast(h = 12)
forecast_values <- vector()
forecast_values <- c(forecast_values, rep(NA, 48))
forecast_values[37:48] <- forecasts$mean[1:12]
plotting_dataset <- Country_level
plotting_dataset$Forecasts <- forecast_values
names(plotting_dataset)[2] <- 'Actuals'
ggplot(plotting_dataset, aes(x = Date)) +
geom_line(aes(y = Actuals)) +
geom_line(aes(y = Forecasts), color = 'Red', linetype = 'dashed', size = 1)## Warning: Removed 36 rows containing missing values (geom_path).
forecasts## Point Forecast Lo 80 Hi 80 Lo 95 Hi 95
## Jan 2015 691.4167 510.3547 872.4787 414.5063 968.3271
## Feb 2015 643.4167 462.3547 824.4787 366.5063 920.3271
## Mar 2015 776.4167 595.3547 957.4787 499.5063 1053.3271
## Apr 2015 750.4167 569.3547 931.4787 473.5063 1027.3271
## May 2015 1131.4167 950.3547 1312.4787 854.5063 1408.3271
## Jun 2015 1498.4167 1317.3547 1679.4787 1221.5063 1775.3271
## Jul 2015 952.4167 771.3547 1133.4787 675.5063 1229.3271
## Aug 2015 1213.4167 1032.3547 1394.4787 936.5063 1490.3271
## Sep 2015 1536.4167 1355.3547 1717.4787 1259.5063 1813.3271
## Oct 2015 983.4167 802.3547 1164.4787 706.5063 1260.3271
## Nov 2015 1518.4167 1337.3547 1699.4787 1241.5063 1795.3271
## Dec 2015 1485.4167 1304.3547 1666.4787 1208.5063 1762.3271## app.R ##
library(shiny)
library(shinydashboard)##
## Attaching package: 'shinydashboard'## The following object is masked from 'package:graphics':
##
## boxsidebar <- dashboardSidebar(
###side bar panel options
sidebarMenu(
menuItem("Exploratory Data Analysis",tabName = "eda"),
menuItem("Forecast",tabName = "forecast"),
menuItem(selectInput("country_id", "Select the Country:",
c("United States","Australia","Mexico","Germany","China","France"
)))
))
body <- dashboardBody(
tabItems(tabItem(tabName = "eda",
fluidRow(
box( width=12, plotOutput("plot1"))
),
fluidRow(
box( width=12, plotOutput("plot2"))),
fluidRow(box( width=12, plotOutput("plot3")))),
tabItem(tabName = "forecast",
fluidRow(
box(width =12, plotOutput("plot4"))),
fluidPage(
sliderInput("period", "Forecast Period (in months):",
min = 12, max = 60, step=12,value = 24),plotOutput("plot5")))))
ui <- dashboardPage(skin = "red",
dashboardHeader(title="Sales Forecasting"),
sidebar,
body
)
server <- function(input, output) {
#monthly_ts
output$plot1<-renderPlot({
Country_picked <- sales[which(Country_filtered$Country %in% c(input$country_id)),]
Country_level <- aggregate(cbind( Quantity) ~ Date, data = Country_picked, FUN = sum)
ggplot(Country_level, aes(y = Quantity, x = Date)) +
scale_x_datetime(labels = date_format("%b"), date_breaks = '1 month', expand = c(0.05,0.1)) +
facet_grid(~year(Date), space = "free_x", scales = "free_x", switch='x')+
geom_point() + geom_line(linetype = 'dashed') +
ggtitle("Monthly Time Series Plot") +
theme_bw() +
theme(
plot.title = element_text(hjust = 0.5),
strip.placement = 'outside',
strip.background = element_rect( colour = "grey50"),
#panel.spacing = unit(0,'cm'),
axis.text.x = element_text(angle = 90)
)
})
##tab_2
#cat_ts
output$plot2<-renderPlot({
Country_picked <- sales[which(Country_filtered$Country %in% c(input$country_id)),]
Category_sales <- aggregate(cbind(Quantity) ~ Category + Date, data = Country_picked, FUN = sum)
ggplot(Category_sales, aes(y = Quantity, x = Date)) +
scale_x_datetime(labels = date_format("%b"), date_breaks = '1 month', expand = c(0.05,0.1)) +
facet_grid(~year(Date), space = "free_x", scales = "free_x", switch='x')+
geom_point() + geom_line( aes(color = Category,linetype = Category), size = 1) +
ggtitle("Monthly Time Series Plot by Category") +
theme_bw() +
theme(
plot.title = element_text(hjust = 0.5),
strip.placement = 'outside',
strip.background = element_rect( colour = "grey50"),
#panel.spacing = unit(0,'cm'),
axis.text.x = element_text(angle = 90)
)
})
#heatmap
output$plot3 <- renderPlot({
Country_picked <- sales[which(Country_filtered$Country %in% c(input$country_id)),]
Sub_Category_sales <- aggregate(cbind(Quantity) ~ Category + `Sub-Category` + Date, data = Country_picked, FUN = sum)
ggplot(Sub_Category_sales, aes(y = `Sub-Category`, x = Date, fill = Quantity)) +
scale_x_datetime(labels = date_format("%b"), date_breaks = '1 month', expand = c(0,0)) +
facet_grid(~year(Date), space = "free_x", scales = "free_x", switch='x') +
geom_tile() +
geom_text(aes(label = Quantity), size = 3) +
scale_fill_gradient(low="grey", high="red") +
ggtitle("Monthly Time Series Plot by Sub-Category") +
theme_bw() +
theme(
plot.title = element_text(hjust = 0.5),
strip.placement = 'outside',
strip.background = element_rect( colour = "grey50"),
panel.spacing = unit(0,'cm'),
axis.text.x = element_text(angle = 90)
)
})
#forecast plot
output$plot4 <- renderPlot({
Country_picked <- sales[which(Country_filtered$Country %in% c(input$country_id)),]
Country_level <- aggregate(cbind( Quantity) ~ Date, data = Country_picked, FUN = sum)
Sales_train_ts <- ts(Country_level$Quantity[1:36], frequency = 12, start = c(2012,1), end = c(2014,12))
## Country models
if (input$country_id == "United States") {
fit <- Sales_train_ts %>% Arima( order = c(0,1,1), seasonal = c(0,1,0), include.constant = FALSE)
} else
if (input$country_id == "Australia"){
fit <- Sales_train_ts %>% Arima( order = c(1,1,1), seasonal = c(0,1,1), include.constant = TRUE)
}else
if (input$country_id == "France"){
fit <- Sales_train_ts %>% Arima( order = c(0,0,0), seasonal = c(1,1,0), include.constant = TRUE)
}else
if(input$country_id == "China"){
fit <- Sales_train_ts %>% Arima( order = c(1,1,2), seasonal = c(0,1,1), include.constant = FALSE, lambda = 0)
}else
if (input$country_id == "Mexico"){
fit <- Sales_train_ts %>% Arima( order = c(1,0,1), seasonal = c(1,1,0), include.constant = TRUE , lambda = 0)
}else
if (input$country_id == "Germany"){
fit <- Sales_train_ts %>% Arima( order = c(0,1,1), seasonal = c(0,1,1), include.constant = FALSE, lambda = 0)
}else {fit <- Sales_train_ts %>% Arima( order = c(1,1,2), seasonal = c(0,1,1), include.constant = FALSE, lambda = 0)
}
forecasts <- fit %>% forecast(h = 12)
forecast_values <- vector()
forecast_values <- c(forecast_values, rep(NA, 48))
forecast_values[37:48] <- forecasts$mean[1:12]
plotting_dataset <- Country_level
plotting_dataset$Forecasts <- forecast_values
names(plotting_dataset)[2] <- 'Actuals'
ggplot(plotting_dataset, aes(x = Date)) +
scale_x_datetime(labels = date_format("%b"), date_breaks = '1 month', expand = c(0.05,0.1)) +
facet_grid(~year(Date), space = "free_x", scales = "free_x", switch = 'x')+
geom_point(aes(y = Actuals)) +
geom_line(aes(y = Actuals, colour = 'Actuals')) +
geom_line(aes(y = Forecasts, colour = 'Forecasts'), linetype = 'dashed', size = 1) +
scale_color_manual(name = 'Values', values = c('Actuals' = 'Black',
'Forecasts' = 'red')) +
ggtitle("Forecast Comparison") +
theme_bw() +
theme(
plot.title = element_text(hjust = 0.5),
strip.placement = 'outside',
strip.background = element_rect( colour = "grey50"),
panel.spacing = unit(0,'cm'),
axis.text.x = element_text(angle = 90))
})
output$plot5 <- renderPlot({
Country_picked <- sales[which(Country_filtered$Country %in% c(input$country_id)),]
Country_level <- aggregate(cbind( Quantity) ~ Date, data = Country_picked, FUN = sum)
Sales_train_ts <- ts(Country_level$Quantity[1:36], frequency = 12, start = c(2012,1), end = c(2014,12))
## Country models
if (input$country_id == "United States") {
fit <- Sales_train_ts %>% Arima( order = c(0,1,1), seasonal = c(0,1,0), include.constant = FALSE)
} else
if (input$country_id == "Australia"){
fit <- Sales_train_ts %>% Arima( order = c(1,1,1), seasonal = c(0,1,1), include.constant = TRUE)
}else
if (input$country_id == "France"){
fit <- Sales_train_ts %>% Arima( order = c(0,0,0), seasonal = c(1,1,0), include.constant = TRUE)
}else
if(input$country_id == "China"){
fit <- Sales_train_ts %>% Arima( order = c(1,1,2), seasonal = c(0,1,1), include.constant = FALSE, lambda = 0)
}else
if (input$country_id == "Mexico"){
fit <- Sales_train_ts %>% Arima( order = c(1,0,1), seasonal = c(1,1,0), include.constant = TRUE , lambda = 0)
}else
if (input$country_id == "Germany"){
fit <- Sales_train_ts %>% Arima( order = c(0,1,1), seasonal = c(0,1,1), include.constant = FALSE, lambda = 0)
}else {fit <- Sales_train_ts %>% Arima( order = c(1,1,2), seasonal = c(0,1,1), include.constant = FALSE, lambda = 0)
}
plot(forecast(fit,h=input$period ))
})
}
shinyApp(ui, server)## PhantomJS not found. You can install it with webshot::install_phantomjs(). If it is installed, please make sure the phantomjs executable can be found via the PATH variable.Shiny applications not supported in static R Markdown documents