---
title: "IBM Data Science Experience"
runtime: shiny
output: 
  flexdashboard::flex_dashboard:
    orientation: rows
    vertical_layout: fill
    source_code: https://github.com/IBMDataScience/SparkSummitDemo
    social: [ "twitter", "facebook", "linkedin" ]

    
---

Data Exploration 
=====================================

```{r setup, include=FALSE}


### Run this first before Knitting the RMD ##############
if(!require(devtools)){
    install.packages("devtools")
    library(devtools)
}

if(!require(flexdashboard)){
    install_github("gfilla/flexdashboard")
    library(flexdashboard)
}

########################################################

#check for packages install if necessary
packages <- function(x){
  x <- as.character(match.call()[[2]])
  if (!require(x,character.only=TRUE)){
    install.packages(pkgs=x,repos="http://cran.r-project.org")
    require(x,character.only=TRUE)
  }
}
packages(shiny)
packages(leaflet)
packages(rbokeh)
packages(ggplot2)
packages(knitr)
packages(dplyr)
packages(plotly)
packages(xts)
packages(dygraphs)
packages(png)
packages(RCurl)
library(devtools)
library(shiny)
library(flexdashboard)
library(leaflet)
library(rbokeh)
library(ggplot2)
library(knitr)
library(dplyr)
library(plotly)
library(xts)
library(dygraphs)
library(png)
library(RCurl)

#setwd('/srv/shiny-server/shinyDemo')
setwd('~/')#home 

datdf <- getURL('https://raw.githubusercontent.com/IBMDataScience/SparkSummitDemo/master/data/clusterEnergyLocation.csv')
df <- read.csv(text = datdf)

#Jitter buildings that are at same lat/long
df$Lat <- round(jitter(df$Lat, factor = 0.00001),5)
df$Long<- round(jitter(df$Long, factor = 0.00001),5)


df$kmeans_label <- df$kmeans_label + 1
df$kmeans_label <- as.factor(df$kmeans_label)
df$ReClustering <- df$ReClustering+1
df$ReClustering <- as.factor(df$ReClustering)

df$plugChg <-mean(df$plug_load_consumption)/df$plug_load_consumption  - 1
df$acChg <- mean(df$ac_consumption)/df$ac_consumption - 1
df$domesticChg <-mean(df$domestic_gas)/df$domestic_gas - 1
df$heatingChg <-mean(df$heating_gas)/df$heating_gas - 1
df$EnergySurplus <- round(((1*(df$plugChg + df$acChg + df$domesticChg + df$heatingChg))/4)*100,0)

df$billValues <- gsub('\\$','', df$Annual.Energy.Bill..USD.)
df$billValues <- gsub(',','', df$billValues)
df$billValues <- as.numeric(gsub(' ','', df$billValues))

dat <- getURL('https://raw.githubusercontent.com/IBMDataScience/SparkSummitDemo/master/data/CombinedSensorData.csv')
allSensors <- read.csv(text = dat)

allSensors$Unitarian.s <- as.POSIXct(allSensors$Unitarian.s,format='%m/%d/%Y %H:%M')
temps <- data.frame(allSensors$Unitarian.s, allSensors$Unitarian.t, allSensors$Parish.t, allSensors$Basement.t,allSensors$Office.t)
colnames(temps) <- c('Time','SanctuaryTemp','ParishTemp','BasementTemp','OfficeTemp')
t<- temps[,-1]
allTemps <- xts(t,order.by = temps$Time)

#function used for scoring
scoreProperty <- function(num_stories,sq_feet, num_plugged_equipment){
    # b's come from regression model
    # n's come from normalization
    b0 = -0.030743500313
    b2 = 0.0285973339538 
    b3 = 0.668102284559 
    b4 = 0.374074888107
    
    n0 = 380400.00
    n2 = 4.00
    n3 = 65000.00
    n4 = 43.00

    x2 = num_stories/n2
    x3 = sq_feet/n3
    x4 = num_plugged_equipment/n4
    energy = b0 + b2 * x2 + b3 * x3 + b4 * x4
    kWh = energy*n0
    usd = kWh*0.18 # $0.18 per kWh in NYC (on average)
    return(c(kWh, usd))
  }

bplogo <- "https://raw.githubusercontent.com/IBMDataScience/SparkSummitDemo/master/blocpowertransp.png"

```

Row {data-height=200}
-----------------------------------------------------------------------

### ![blocpower](`r bplogo`) Build a Better World.  Develop green energy projects in American inner cities.

```{r}
renderValueBox({
  
  valueBox(value = '', icon = "")
})

```

### Annual Energy Bill (USD)

```{r}
renderValueBox({
  bill <- prop_data()
  #cat(toString(bill))
  cat(toString(getwd()))
  valueBox(value = bill$Annual.Energy.Bill..USD., icon = "glyphicon-usd")
})
```

### Average Energy Compared to Benchmark

```{r}
renderValueBox({
  energyUse <- prop_data()$EnergySurplus
  valueBox(value = paste0(energyUse,'%'), 
           icon = "glyphicon-tree-deciduous",
           color =   ifelse(energyUse < 0, "danger", "success")
           )
})
```



Row {data-height=600}
-----------------------------------------------------------------------

### Energy Consumption Map 

```{r}

click_marker <- eventReactive(input$map_marker_click, {

    x <- input$map_marker_click
    
    y <-c(x[3], x[4])
    #cat(toString(y))
    return(y)
  })


qpal <- colorQuantile("YlGnBu", df$Measured)

output$map <- renderLeaflet({

    map2 <- leaflet(data = df) %>% 
      addProviderTiles("CartoDB.Positron")%>% 
      addCircleMarkers(
          radius = 6,
          fillColor = ~qpal(df$Measured),fillOpacity = 0.7,
          stroke = T, weight =2,color='#2b1d0e',
          popup = paste('Property: ', df$property_name,"<br> Annual Bill: ", df$Annual.Energy.Bill..USD.)) %>% 
      addLegend(position = "bottomright",na.label = "NA", title="Annual Energy Cost",colors=c('#FFFFD9','#99D6B9','#2280B8','#081D58'), labels = c('< $2,000','$2,000 - $4,000','$4,001- $6,000','$6,000 +'))
    map2
})



prop_data <- reactive({
  
  clicked_building <- click_marker()
  # Fetch data for the clicked tract
    barchart_values <- df[which(df$Lat == round(as.numeric(clicked_building[1]),5) & df$Long == round(as.numeric(clicked_building[2]),5)),]
  # barchart_values <- df[which(df$Lat == as.vector(clicked_building[1]) & df$Long == as.vector(clicked_building[2])),]
cat(toString(barchart_values))
  return(barchart_values)

})

leafletOutput('map')  

```

### Selected Building Energy Types

```{r}

output$bars <- renderPlotly({
  barplot_data <- prop_data()
  #cat(toString(barplot_data))
  
  xLabel <- list(
  title = " "
 )
  yLabel <- list(
  title = "Energy Consumption",
  showticklabels = F,
  range = list(0,1)
) 
  
  
  if(dim(barplot_data)[1] != 0){
    
    #print(barplot_data)
        updated_vals <- c(barplot_data$plug_load_consumption, barplot_data$ac_consumption, barplot_data$domestic_gas, barplot_data$heating_gas)
      
      p <- plot_ly(
        x = c('Plug','AC','Home Gas', 'Heat Gas'),
        y = updated_vals,
        name = "Selected Property Values",
        type = "bar") %>% 
        layout(xaxis = xLabel, yaxis=yLabel)
     p 
  }else{
      bar_vals <- c(mean(df$plug_load_consumption), mean(df$ac_consumption), mean(df$domestic_gas), mean(df$heating_gas))
    
        p <- plot_ly(
            x =c('Plug','AC','Home Gas', 'Heat Gas'),
            y = bar_vals,
            name = "Mean Property Values",
            type = "bar") %>% 
            layout(xaxis = xLabel, yaxis=yLabel)
       p
  }
  
    p2 <- add_trace(
      p,
      x =c('Plug','AC','Home Gas', 'Heat Gas'),
      y = c(mean(df$plug_load_consumption),mean(df$ac_consumption), mean(df$domestic_gas),mean(df$heating_gas)),
      name = "Mean Values",
      type = "bar")
  p2
  
})  

plotlyOutput('bars')

```


Row {data-height=300}
-----------------------------------------------------------------------


### Sensor Temperature Data - Different Building Locations

```{r}
  
dygraph(allTemps)


```


Cluster Analysis
=====================================


Row
-----------------------------------------------------------------------

### Energy Consumption Map

```{r}


# qualPalette <- c("#d7191c", "#fdae61", "#a6d96a", "#1a9641")
# clusterLabels <- seq(1:4)
# df$color <- qualPalette[match(df$kmeans_label, clusterLabels)]
# qpal2 <- colorFactor(palette=qualPalette, df$kmeans_label)

qualPalette <- c("#1a9641","#d7191c")
clusterLabels <- seq(1:2)
df$color <- qualPalette[match(df$ReClustering, clusterLabels)]
qpal2 <- colorFactor(palette=qualPalette, df$ReClustering)


  map2 <- leaflet(data = df) %>% 
    
  addProviderTiles("CartoDB.Positron")%>% 
  addCircleMarkers(
    lng = df$Long,
    lat = df$Lat,
    radius = 6,
    color = ~qpal2(df$ReClustering),
    stroke = F, fillOpacity = 0.7,
    popup = ~property_name)

addLegend(map2,position = "bottomright", colors=qualPalette,labels=c("Efficient","Inefficient"),na.label = "NA", title="Cluster Labels", opacity =0.7)
map2  


```

### Clusters by Heating and Plug Consumption

```{r}

p2 <- figure() %>%
  ly_points(heating_gas*100, plug_load_consumption*100, data = df,fill_color =color,fill_alpha = 0.6, hover = c(property_name,Annual.Energy.Bill..USD. ), xlab='Heating Gas', ylab = 'Plug Load Consumption') %>% 
y_axis(number_formatter = "printf", format = "%d%%") %>% 
x_axis(number_formatter = "printf", format = "%d%%")
p2



```

Predictive Model
=====================================

Row  {.sidebar data-width=400}
-----------------------------------------------------------------------

### Predict Energy Use and Cost for New Property

```{r}

#coefficient 1 - Stories
sliderInput("stories", "Enter number of Stories:",
            min = 0, max = 4, value = 1, step = 1
)
#coefficient 2 - Sq. Foot
sliderInput("sqft", "Enter Property Square Footage:",
            min = 0, max = 65000, value = 20000, step = 100
)

#coefficient 3 - # of Plugged Outlets
sliderInput("plugged", "Enter Number of Plugged Outlets:",
            min = 0, max = 43, value = 29, step = 1
)


score <- reactive({
  scoreProperty(input$stories,input$sqft,input$plugged)
})

#cat(toString(score[1]))
 

```

Row {data-height=200}
-----------------------------------------------------------------------



### Predicted Annual Energy Bill (USD @ $0.18/kWh)

```{r}
renderGauge({
  bill<- score()[2]
  gauge(bill, min = 0, max = 100000, symbol = '$', gaugeSectors(
  success = c(0, 2000), warning = c(2001, 15000), danger = c(15001, 100000)))
  
})
  
```

### Predicted Annual Energy (kWh)

```{r}

renderGauge({
  kwh <- score()[1]
  gauge(kwh, min = 0, max = 380400, gaugeSectors(
 success = c(0, 100000), warning = c(100001,200000), danger = c(200001,400000)))
  
})

```


Row 
-----------------------------------------------------------------------


### Accuracy of Linear Regression Fit

```{r}

 p3 <- figure() %>%
ly_points(Measured ,Predicted, data = df,fill_color = color, fill_alpha=0.6,hover = c(property_name, ReClustering), xlab='Measured Energy Usage', ylab='Predicted Energy Usage') 
p3<- ly_abline(p3,a=0, b=1, color='blue')
p3


```

Data Table
=====================================


Column 
-----------------------------------------------------------------------

### Raw Data Table

```{r}

  kable(df)



```