library(shiny)
library(e1071)
library(rminer)
library(dplyr)
library(tidyr)
library(ggplot2)
library(ggvis)
library(corrplot)
library(DT)
library(caret)
ui <- navbarPage(title = "HR Analytics         ",

                 tabPanel("Data Import",
                          sidebarLayout(sidebarPanel(
                            fileInput('file1', 'Choose CSV File to upload',
                                      accept=c('text/csv',
                                               'text/comma-separated-values,text/plain',
                                               '.csv')),
                            helpText("Note: Please ensure that the the file is in .csv",
                                     "format and contains headers."),
                            tags$hr(),
                            actionButton("do", "Import")
                          ),
                          mainPanel(h2(helpText("Descriptive Statistics")),
                                    verbatimTextOutput('contents'))
                          )
                 ),#tabpanel
                 tabPanel("Predictive Model",
                          sidebarLayout(sidebarPanel(
                            uiOutput("model_select"),
                            actionButton("enter", "Enter")
                          ),
                          mainPanel(h2(helpText("Model Output")),
                                    verbatimTextOutput('modelOutput'))
                          )
                 ),#tabpanel
                 tabPanel("Report",
                          sidebarLayout(sidebarPanel(
                            tags$style(type="text/css",
                                       ".shiny-output-error { visibility: hidden; }",
                                       ".shiny-output-error:before { visibility: hidden; }"
                            ),
                            helpText("Download final list of employess to be retained"),
                            br(),
                            uiOutput("modsel"),
                            helpText("Select Model"),
                            uiOutput("noselect"),
                            helpText("Select number pf employess"),
                            downloadButton('downloadData', 'Download'),
                            helpText("Download final list of employees to be retained")
                          ),

                          mainPanel(h2(helpText("Retained Employees")),
                                    dataTableOutput("reportOutput"))
                          )
                 )#tabpanel


)


library(shiny)

server <- function(input, output) {

  hr = eventReactive(input$do,{
    inFile <- input$file1

    if (is.null(inFile))
      return(NULL)

    hr = read.csv(inFile$datapath, header=T, sep=",")
  })

  output$contents <- renderPrint({
    return(summary(hr()))
  })

  output$model_select<-renderUI({
    selectInput("modelselect","Select the model",choices = c("Tree Learning"="rpart","Logistic Regression"="LogitBoost", "Naive Bayes" = "nb"))
  })

  output$modsel<-renderUI({
    selectInput("modelselect2","Select Algo",choices = c("Logistic Regression","Naives Bayes","Tree Learning"),selected = "Logistic_reg")
  })

  output$noselect<- renderUI({
    sliderInput("noselect", "Number of observations:",
              min = 0, max = 300, value = 20)})


  algo = eventReactive(input$enter,{
    return(input$modelselect)
  })


  output$modelOutput <- renderPrint({
    hr_model <- hr() %>% filter(left==0 | last_evaluation >= 0.70 | time_spend_company >= 4 | number_project > 5)
    hr_model$left <- as.factor(hr_model$left)
    train_control<- trainControl(method="cv", number=5, repeats=3)
    rpartmodel<- train(left~., data=hr_model, trControl=train_control, method=algo())
    # make predictions
    predictions<- predict(rpartmodel,hr_model)
    hr_model_tree<- cbind(hr_model,predictions)
    # summarize results
    confusionMatrix<- confusionMatrix(hr_model_tree$predictions,hr_model_tree$left)
    confusionMatrix
  })

  rt <- reactive(
    if(input$modelselect2== "Logistic Regression"){
      f1<-data()
      hr_model1 <- hr() %>% filter(left==0 |last_evaluation >= 0.70 | time_spend_company >= 4 | number_project > 5)
      hr_model1$left <- as.factor(hr_model1$left)
      train_control<- trainControl(method="cv", number=5, repeats=3)
      # Keep some data to test again the final model
      set.seed(100)
      inTraining <- createDataPartition(hr_model1$left, p = .75, list = FALSE)
      training <- hr_model1[ inTraining,]
      testing  <- hr_model1[-inTraining,]
      # Estimate the drivers of attrition
      logreg = glm(left ~ ., family=binomial(logit), data=training)
      # Make predictions on the out-of-sample data
      probaToLeave=predict(logreg,newdata=testing,type="response")
      # Structure the prediction output in a table
      predattrition = data.frame(probaToLeave)
      # Add a column to the predattrition dataframe containing the performance
      predattrition$performance=testing$last_evaluation

      predattrition$priority=predattrition$performance*predattrition$probaToLeave
      orderpredattrition=predattrition[order(predattrition$priority,decreasing = TRUE),]
      orderpredattrition <- head(orderpredattrition, n=input$noselect)
      or<- data.frame(orderpredattrition)
      or
    }
    else if(input$modelselect2== "Naives Bayes"){
      f1<-data()
      hr_model1 <- hr() %>% filter(left==0 |last_evaluation >= 0.70 | time_spend_company >= 4 | number_project > 5)
      hr_model1$left <- as.factor(hr_model1$left)
      train_control<- trainControl(method="cv", number=5, repeats=3)
      # Keep some data to test again the final model
      set.seed(100)
      inTraining <- createDataPartition(hr_model1$left, p = .75, list = FALSE)
      training <- hr_model1[ inTraining,]
      testing  <- hr_model1[-inTraining,]

      # Estimate the drivers of attrition
      e1071model2 = naiveBayes(left ~ ., data=training)
      # Make predictions on the out-of-sample data
      probaToLeave=predict( e1071model2,newdata=testing[,c(-7,-9,-10)],type="raw")
      # Structure the prediction output in a table
      predattrition = data.frame(probaToLeave)
      colnames(predattrition) <- c("c","probaToLeave")
      predattrition[1] <- NULL
      # Add a column to the predattrition dataframe containing the performance
      predattrition$performance=testing$last_evaluation
      predattrition$priority=predattrition$performance*predattrition$probaToLeave
      orderpredattrition=predattrition[order(predattrition$priority,decreasing = TRUE),]
      orderpredattrition <- head(orderpredattrition, n=input$noselect)
      or<- data.frame(orderpredattrition)

    }

    else if(input$modelselect2== "Tree Learning"){
      f1<-data()
      hr_model1 <- hr() %>% filter(left==0 |last_evaluation >= 0.70 | time_spend_company >= 4 | number_project > 5)
      hr_model1$left <- as.factor(hr_model1$left)
      train_control<- trainControl(method="cv", number=5, repeats=3)
      # Keep some data to test again the final model
      set.seed(100)
      inTraining <- createDataPartition(hr_model1$left, p = .75, list = FALSE)
      training <- hr_model1[ inTraining,]
      testing  <- hr_model1[-inTraining,]
      # Estimate the drivers of attrition
      rpartmodel = rpart(left ~ satisfaction_level+last_evaluation+number_project+average_montly_hours+time_spend_company+Work_accident+promotion_last_5years,method = "anova",data=training)
      # Make predictions on the out-of-sample data
      probaToLeave=predict(rpartmodel,newdata=testing[,c(-7,-9,-10)],type="vector")
      # Structure the prediction output in a table
      predattrition = data.frame(probaToLeave)*0.5
      # Add a column to the predattrition dataframe containing the performance
      predattrition$performance=testing$last_evaluation

      predattrition$priority=predattrition$performance*predattrition$probaToLeave
      orderpredattrition=predattrition[order(predattrition$priority,decreasing = TRUE),]
      orderpredattrition <- head(orderpredattrition, n=input$noselect)

      or<- data.frame(orderpredattrition)
      or
    }
  )


  output$reportOutput = renderDataTable({
    rt()
  })

  output$downloadData <- downloadHandler(
    filename = function() { paste(input$modelselect2, '.csv', sep='') },
    content = function(file){
      write.csv(rt(), file)
    }
  )

}
shinyApp(ui=ui, server = server)

output$table_out  <- DT::renderDataTable(
                datatable(
                    data,
                rownames = TRUE,
                options = list(
                    fixedColumns = TRUE,
                    autoWidth = TRUE,
                    ordering = FALSE,
                    dom = 'tB',
                    buttons = c('copy', 'csv', 'excel', 'pdf')
                ),
                class = "display" #if you want to modify via .css
            )

DT::dataTableOutput("table_out")