---
title: "UCI Machine Learning datasets for R"
author: "Andrzej Wójtowicz"
output: 
  html_document: 
    keep_md: yes
---

```{r global-options, include=FALSE}
knitr::opts_chunk$set(comment = "", echo = FALSE, warning = FALSE, message = FALSE)
source('init.R')
```

Document generation date: `r Sys.time()`.

This project preprocesses a few datasets from [UC Irvine Machine Learning
Repository](https://archive.ics.uci.edu/ml/) into tidy R object files.
It focuses on the binary classification datasets and saves only complete cases
within a dataset.

**R software**: [Microsoft R Open](https://mran.microsoft.com/open/) (`r paste0(R.version$major, ".", R.version$minor)`)

**Reproducibility library**: [checkpoint](https://github.com/RevolutionAnalytics/checkpoint)

**Reproducibility procedure**:

 1. Run *s1-download-data.R* to download original datasets.
 2. Run *s2-preprocess-data.R* to preprocess the datasets.
 
 Optionally:
 
 3. knit *s3-make-readme.Rmd* to get an overview of the preprocessed datasets,
 4. run *s4-make-release.sh* to create zip file with preprocessed datasets.

```{r show-datasets, results='asis'}

cat("\n# Table of Contents\n\n")

for (dir.name in dir(DATASETS.DIR))
{
    config.yaml.file.path = file.path(DATASETS.DIR, dir.name, DATASET.CONFIG.FILE)
    config.yaml = yaml.load_file(config.yaml.file.path)
    
    anchor = gsub(" ", "-", gsub("[[:punct:]]", "", 
                                 tolower(config.yaml$name)))
    cat(paste0("1. [", config.yaml$name, "](#", anchor, ")\n" ))
}

cat("\n---\n\n")

for (dir.name in dir(DATASETS.DIR))
{
    config.yaml.file.path = file.path(DATASETS.DIR, dir.name, DATASET.CONFIG.FILE)
    config.yaml = yaml.load_file(config.yaml.file.path)
    
    cat(paste("#", config.yaml$name, "\n\n"))
    
    cat(paste("**Local directory**:", dir.name, "\n\n"))
    
    cat(paste0("**Details**: [link](", config.yaml$info, ")\n\n"))
    
    cat(paste("**Source data files**:\n\n"))
    for (file.url in config.yaml$urls)
    {
        cat(paste0("* [", URLdecode(basename(file.url)), "](", file.url, ")\n"))
    }
    cat("\n")
    
    cat(paste0("**Cite**:\n```nohighlight\n", config.yaml$cite, "\n```\n\n"))
    
    cat(paste("**Dataset**:\n\n"))
    
    preprocessed.dir =       gsub(DATASET.NAME.PATTERN, dir.name,
                                  DATASET.PREPROCESSED.DIR)
    preprocessed.file.path = file.path(preprocessed.dir,
                                       DATASET.PREPROCESSED.OUTPUT.FILE)
    
    dataset = readRDS(preprocessed.file.path)
    
    cat("```nohighlight\n")
    cat(str(dataset))
    cat("\n```\n\n")
    
    cat("**Predictors**:\n\n")
    
    df.pred = data.frame(table(sapply(dataset[, 1:(ncol(dataset) - 1)], 
                               function(f){paste(class(f), collapse = " ")})))
    colnames(df.pred) = c("Type", "Frequency")
    
    cat(knitr::kable(df.pred, format = "markdown"), sep = "\n")
    cat("\n")
    
    perc.classes = sort(round(100*as.numeric(
        table(dataset[, ncol(dataset)]))/nrow(dataset), 0))
    num.classes = sort(as.numeric(table(dataset[, ncol(dataset)])))
    
    cat("**Class imbalance**:\n\n")
    
    cat(knitr::kable(data.frame(A = c(paste(perc.classes[1], "%"), 
                                      num.classes[1]), 
                                B = c(paste(perc.classes[2], "%"),
                                      num.classes[2])), 
                     format = "markdown", col.names = c("class A", "class B"),
                     align = c("c", "c")), 
        sep = "\n")

    cat("\n---\n\n")
}
```