nearest_neighbour.ex

defmodule KNN do
  defmodule LabeLWithFeatures do
    defstruct label: [], pixels: []
  end

  def slurp_file(file) do
    File.read!(file)
    |> String.split("\n")
    |> Stream.drop(1)
    |> Stream.map(fn row ->
      csv_row_to_label(String.split(row, ","))
    end)
    |> Enum.to_list()
  end

  def csv_row_to_label([label|pixels]) do
    %LabeLWithFeatures{label: label, pixels: pixels |> Enum.map(fn num -> parse_binary_to_float(num) end)}
  end

  def parse_binary_to_float(binary) do
    {float, _} = Integer.parse(binary)
    float
  end

  def distance(x_pixels, y_pixels) do
    x_pixels
    |> Stream.flat_map(fn x -> (Stream.map(y_pixels, fn y -> :math.pow(x-y, 2) end )) end)
    |> Enum.to_list()
    |> Enum.sum
  end

  def classify(training_labels, pixels) do
    min_training_label =
    training_labels
    |> Enum.min_by(fn training_label -> distance(training_label.pixels, pixels) end)
    min_training_label.label
  end

  def main() do
    trainingsample = slurp_file("trainingsample.csv")
    validationsample = slurp_file("validationsample.csv")

    total_correct_item =
    validationsample
    |> Stream.filter(fn validation_item -> (classify(trainingsample, validation_item.pixels) == validation_item.label) end)
    |> Enum.to_list()
    |> Enum.count()

    IO.puts "Percentage correct: #{(total_correct_item/Enum.count(validationsample)) * 100}"
  end
end

KNN.main()
# IO.inspect NearestNeighbour.slurp_file("trainingsample.csv") |> hd()
# IO.puts NearestNeighbour.calculate_total_correct("trainingsample.csv", "validationsample.csv")