CODEWITHSUNDEEP
rrstudioshiny
Siddharth Pandit
Siddharth Pandit

Reputation: 27

Executing code on data set filtered using drop down menu in Shiny/R

I am trying to create a shiny app to accomplish the following -

  1. Select an ID from the drop down menu (Eg: 106841)
  2. Filter the original data set of about 150K observations and obtain a smaller data set of about 600 values using the primary key
  3. Perform additional manipulation and run regression on this data set
  4. Present ggplot graph and text summary of the regression
  5. Allow same process to be applied to different IDs based on user input

Sample structure of code to give you an idea of how i am trying to meet these requirements -

Server.R:
#Excerpt of server code
branch_data <- openxlsx::read.xlsx("Branch_Final.xlsx")
<Other data input and cleaning code>
branch_data_final <- data.table(branch_data)

shinyServer(function(input, output) {

#Filtering data set using ID (input$select_ID is the variable)
data_branch_analysis<-(data_branch_analysis[ID==input$select_ID])[order(DATE)]

#Data manipulation for regression
data_branch_analysis[,NDATE:=as.Date(DATE,"%Y.%m.%d")]
data_branch_analysis[,L_AVG_AGE:=shift(AVG_AGE,1)]
data_branch_analysis[,L_AVG_WAGE:=as.numeric(shift(AVG_WAGE,1))]
<Other lines of code for manipulation>

#Regression
 fit1<-lm(data=data_branch_analysis,VISITOR_NUM~0+time+WD+L_W3_7+L_W7_14+L_W14_21+...)
bestm<-step(fit1)
fit2<-auto.arima(data_branch_analysis_train$VISITOR_NUM,max.order=30,xreg=as.matrix(x_reg))
<Other lines of code for regression)

#GGPLOT 

  output$final_forecast_branch <- renderPlot({
    g <-
      ggplot()+geom_line(aes(x=data_branch_analysis$NDATE,y=data_branch_analysis$VISITOR_NUM,col="original"))+
      geom_line(aes(x=data_branch_analysis$NDATE[2:(ntrain+1)],y=fit2$fitted,col="train"))+
     geom_line(aes(x=data_branch_analysis$NDATE[(ntrain+2):nrow(data_branch_analysis)],y=fore2$mean,col="test"))
    g
  })

output$final_forecast_branch_analysis_accuracy <- renderText(expr = accuracy(fore2,x=data_branch_analysis_$VISITOR_NUM[(ntrain+2):nrow(data_branch_analysis)])


}

UI.R
#Excerpt of UI code

navbarMenu("Analyzer Widget",
               tabPanel(
                 "Branch",
                 sidebarLayout(
                   fluid = 'TRUE',
                   sidebarPanel(
                   # p("Please enter the following information - "),
                     selectInput(
                       inputId = "select_ID",
                       'Select Branch ID',
                       selected = "106841",
                       sort(unique(data_branch_analysis$ID))
                     )
                   ),
                   mainPanel(tabsetPanel(
                     tabPanel(
                       'Training Data',
                       plotOutput('final_forecast_branch'),
                       p("Accuracy of Model"),
  textOutput("final_forecast_branch_analysis_accuracy"),
...

Currently, i see no output from the ggplot or textOutput blocks. I've tried reactive and observe but obviously i am unable to implement in properly. Would appreciate your thoughts on how to structure the code for this to work.

Thanks for your inputs.

Upvotes: 0

Views: 641

Answers (1)

Bertil Baron
Bertil Baron

Reputation: 5003

Hi you ewant to make a chain of reactive expressions someting like my example below.

shinyServer(function(input, output) {
  filterDta <- reactive({
    #Filtering data set using ID (input$select_ID is the variable)
    data_branch_analysis<-(data_branch_analysis[ID==input$select_ID])[order(DATE)]

  })
  minpulateDta <- reactive({
    data_branch_analysis <- filterDta()
    #Data manipulation for regression
    data_branch_analysis[,NDATE:=as.Date(DATE,"%Y.%m.%d")]
    data_branch_analysis[,L_AVG_AGE:=shift(AVG_AGE,1)]
    data_branch_analysis[,L_AVG_WAGE:=as.numeric(shift(AVG_WAGE,1))]
    <Other lines of code for manipulation>

  })

  calcRegression <- reactive({
      #Regression
    data_branch_analysis <- minpulateDta()  
      fit1<-lm(data=data_branch_analysis,VISITOR_NUM~0+time+WD+L_W3_7+L_W7_14+L_W14_21+...)
    bestm<-step(fit1)
    fit2<-auto.arima(data_branch_analysis_train$VISITOR_NUM,max.order=30,xreg=as.matrix(x_reg))
    <Other lines of code for regression)

  })

#GGPLOT 

output$final_forecast_branch <- renderPlot({
  data_branch_analysis <- calcRegression()
  g <-
    ggplot()+geom_line(aes(x=data_branch_analysis$NDATE,y=data_branch_analysis$VISITOR_NUM,col="original"))+
    geom_line(aes(x=data_branch_analysis$NDATE[2:(ntrain+1)],y=fit2$fitted,col="train"))+
    geom_line(aes(x=data_branch_analysis$NDATE[(ntrain+2):nrow(data_branch_analysis)],y=fore2$mean,col="test"))
  g
})

output$final_forecast_branch_analysis_accuracy <- renderText(expr = accuracy(fore2,x=data_branch_analysis_$VISITOR_NUM[(ntrain+2):nrow(data_branch_analysis)])


}

In this case of course you don't really need to seperate minpulateDta and calcRegression but it makes the code more readable when you separete the different steps. If you want to reuse the results some where else it is also easier in this way.

Hope this helps!

Upvotes: 1

Related Questions