20180508_OutputBatclassify.Rmd

---
title: "Output of Batclassify"
author: "JF Godeau"
date: "`r Sys.setlocale('LC_ALL', 'en_GB.UTF-8'); format(Sys.time(), '%d %B %Y')`"
output:
  html_document:
    fig_caption: yes
    self_contained: no
    toc: yes
  pdf_document:
    toc: yes
  word_document:
    toc: yes
editor_options: 
  chunk_output_type: console
---

```{r setup, include=FALSE}
#FILE <- file.choose()
DIR <- "/media/jf/Elements/Audiomoth/20180620_Florenville/COSSUS"
FILE <- paste(DIR,"Results.csv",sep="/")
require(data.table)
require(lubridate)
require(knitr)
knitr::opts_chunk$set(echo = FALSE, message = FALSE)
#knitr::opts_knit$set(root.dir=DIR) # ???
# Threshold value of probablilities to discard
Trsh <- 0.69
```

## General info

```{r 'Reformat Date and Time'}
RBC <- read.csv(FILE)
DT.RBC <- data.table(RBC)
DT.RBC[, Date := ymd(substring(DT.RBC$FileName,1,10))]
DT.RBC[, Time := as.character(hms(substring(DT.RBC$FileName,12,19)))]
PER <- max(ymd_hms(DT.RBC$FileName, tz="Europe/Amsterdam")) - min(ymd_hms(DT.RBC$FileName, tz="Europe/Amsterdam"))
PER.h <- round(difftime(max(ymd_hms(DT.RBC$FileName, tz="Europe/Amsterdam")),
                        min(ymd_hms(DT.RBC$FileName, tz="Europe/Amsterdam"))
                        ,units="hours"),1)
DT.num <- data.table(DT.RBC[,5:16], keep.rownames=T)
#DT.num[,Ppyg := NULL] ## !! Remove Ppyg data!!
LogiLine2keep <- apply(DT.num, 1, max) > Trsh
LogiLine2keepNoPyg <- apply(DT.num[,-10], 1, max) > Trsh

TM <- as.POSIXct(DT.RBC$FileName, format="%Y-%m-%d_%H_%M_%S")
TM.TSH <- data.frame(time=TM[LogiLine2keepNoPyg])
TM.TSH$valmax <- apply(DT.num, 1, max)[LogiLine2keepNoPyg]
require(ggplot2)
#ggplot(TM.TSH, aes(time, valmax)) + geom_point()
G1 <- ggplot(data.frame(TM), aes(TM, 0)) +
  geom_point() +
  theme_bw()+
  #  geom_point(data=TM.TSH, aes(time, valmax), col="red",  inherit.aes = F) +
  #  scale_y_continuous(limits=c(min(TM.TSH$valmax), max(TM.TSH$valmax))) +
  geom_histogram(data=TM.TSH, aes(x=time), stat ="bin",  inherit.aes = F, alpha= 0.5, binwidth=1000) +
  #  scale_y_continuous(sec.axis = sec_axis(~., name = "NEW")) +
  scale_x_datetime(date_breaks="2 hours") +
  theme(axis.text.x = element_text(angle = 90, vjust = 1.0, hjust = 1.0))+
  xlab("Time") + ylab("N samples > Threshold")

```

Name of the folder: **`r unique(DT.RBC$FilePath)`**  
Number of files: **`r dim(DT.RBC)[1]`**  
Number of files with at least one probability value above the threshold (defined as `r Trsh`): **`r sum(LogiLine2keep)`**  
Dates recorded for these files: `r paste(unique(DT.RBC$Date), collapse="; ")`  
Time difference between first and last recording: **`r round(PER,1)`** , (= `r PER.h` hours).  

## Distribution of the data in time  
Black dots at y=0 shows the recording period.  

```{r, fig.height=3}
G1
```



## Maximal score obtained for each species
### All species of Batclassify result table

```{r "Global probability for each species"}
MxSp <- sapply(DT.num, max)
#kable(MxSp, caption="General score")
kable(data.frame(t(MxSp)), caption="General score")
```

### Only species with a probability above the threshold (`r Trsh`)  

```{r "Global probability for each species Sup threshold"}
kable(data.frame(t(MxSp[MxSp > Trsh])), caption="Species with p > 0.5")
```


## Number of files where probability is at least above the threshold
### For each species  

```{r 'Select the species, if any, above the threshold for each file'}
CompMax <- cbind(max.col(DT.num[LogiLine2keep,], "first"),
                 max.col(DT.num[LogiLine2keep,], "last"))
Follow <- apply(DT.num[LogiLine2keep,] ,
                1,
                function(x){ord <- order(x, decreasing = T)
                paste(paste(colnames(DT.num)[ord][ x[ord] > Trsh][-1],
                            x[ord][ x[ord] > Trsh][-1]), collapse=" / ")
                })
LogiOneMax <- CompMax[,1] - CompMax[,2] == 0
DFSpMax <- cbind(FileName=as.character(DT.RBC$FileName[LogiLine2keep]),
                 SpeciesMax=names(DT.num)[CompMax[,1]],
                 Prob=apply(DT.num, 1, max)[LogiLine2keep],
                 OtherSp = as.character(Follow))

if(any(!LogiOneMax)){
  Idx <- which(!LogiOneMax)
  if(sum(!LogiOneMax)>1){
    DFSpMax[Idx,2] <- apply(matrix(names(DT.num)[CompMax],
                                   nrow=length(LogiOneMax))[Idx,],
                            1,
                            function(x) paste(x, collapse="-"))
  } else {
    DFSpMax[Idx,2] <- paste(names(DT.num)[CompMax[Idx,]], collapse = "-")
  }
}
DFSpMax <- data.frame(DFSpMax)
kable(sort(table(DFSpMax$SpeciesMax), decreasing = T), col.names = c("Species","N"))
```

### For each species per category of probability  

```{r SepProbCat}
kable(as.data.frame.matrix(table(DFSpMax$SpeciesMax, round(as.numeric(as.character(DFSpMax$Prob)),1))),
      caption="Number of files with each species for probablility categories")

```


## Fully detailed results

```{r 'Final table (PDF or docx)', include=F}


kable(DFSpMax) #If not html output
```


```{r 'Final table', eval=T, include=T, results='asis'}
suppressPackageStartupMessages(library(googleVis))
require(googleVis)
op <- options(gvis.plot.tag='chart') ### to export the chart/table in knitted html
Table <- gvisTable(DFSpMax)
plot(Table)
```

## Full table (any > Trsh)  

```{r 'Full table 05', eval=T, include=T, results='asis'}
op <- options(gvis.plot.tag='chart') ### to export the chart/table in knitted html
Table <- gvisTable(DT.RBC[,-c(1,3,4)])
plot(Table)


```