---
title: "Failure rate vs. date"
author: "Grzegorz Sionkowski"
date: "26 09 2016"
output: html_document
---

```{r packages} 
library(data.table)
```

### Load and process train numeric and date files


```{r}
numf <- fread("../input/train_numeric.csv",
             select = c("Id","Response"),
              showProgress = F)
 
 datef <- fread("../input/train_date.csv",
                select=c(2,25,50,75,100,125,150,175,200,250,300,350,400,450,500,550,600,650,700,750,800,850,900,950,1000,1050,1100,1150),
                showProgress = F)


datef$date <- apply(datef,1, function(x) mean(x,na.rm=TRUE) )

samples <- data.table(date=datef$date,Response=numf$Response)
failures <- samples[samples$Response==1,]

```


### Data analysis


```{r}
bins <- seq(0,1800,300)
samplesX <- hist(samples$date,breaks=bins)
failuresX  <- hist(failures$date,breaks=bins)

dt <- data.table(date=samplesX$mids,samples=samplesX$counts,failures=failuresX$counts,failure_rate=failuresX$counts/samplesX$counts)

plot(x=dt$date,y=dt$failure_rate)


```



```{r}
dt
```

It looks like the failure rate rised up to an unacceptable level and then the assembly line was repaired.