report.Rmd

---
title: "Comparison of nearest neighbors algorithm"
author: "Toby Dylan Hocking"
date: "`r Sys.Date()`"
output: rmarkdown::html_vignette
vignette: >
  %\VignetteIndexEntry{Comparison of nearest neighbors algorithm}
  %\VignetteEngine{knitr::rmarkdown}
  %\VignetteEncoding{UTF-8}
---

```{r setup, include = FALSE}
knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)
```

```{r}
data(spam, package="ElemStatLearn")
data(zip.train, package="ElemStatLearn")
str(spam)
str(zip.train)
is.01 <- zip.train[,1] %in% c(0,1)
data.list <- list(
  spam=list(
    features=as.matrix(spam[, 1:57]),
    labels=ifelse(spam$spam=="spam", 1, 0)),
  zip.train=list(
    features=zip.train[is.01, -1],
    labels=as.integer(zip.train[is.01, 1])))
str(data.list)
n.folds <- 5
results.list <- list()
mean.loss.list <- list()
for(data.name in names(data.list)){
  data.set <- data.list[[data.name]]
  str(data.set)
  stopifnot(all(data.set$labels %in% c(0,1)))
  stopifnot(length(data.set$labels) == nrow(data.set$features))
  set.seed(1)
  fold.vec <- sample(rep(1:n.folds, l=nrow(data.set$features)))
  str(fold.vec)
  result.mat.list <- list()
  for(test.fold in 1:n.folds){
    is.test <- fold.vec == test.fold
    is.train <- !is.test
    X.train <- data.set$features[is.train,]
    y.train <- data.set$labels[is.train]
    baseline <- mean(y.train)
    fit <- nearestNeighbors::NearestNeighborsCV(X.train, y.train, 30L)
    str(fit)
    mean.loss.list[[paste(data.name, test.fold)]] <- fit$mean.loss$mean.loss
    pred.prob.list <- list(
      earlyStopping=.....,
      L2regularized=.....,
      ##knn=fit$predict(data.set$features[is.test,]),
      baseline=rep(baseline, sum(is.test)))
    str(pred.prob.list)
    test.fold.result.list <- list()
    for(algo in names(pred.prob.list)){
      pred.prob.vec <- pred.prob.list[[algo]]
      pred.class.vec <- ifelse(pred.prob.vec > 0.5, 1, 0)
      test.fold.result.list[[algo]] <-
        mean(data.set$labels[is.test] != pred.class.vec)
    }
    result.mat.list[[test.fold]] <- do.call(c, test.fold.result.list)
  }
  do.call(rbind, result.mat.list)
}
```

## Analysis of spam data set



## Analysis of zip.train data set