add preprocessing normalization for data set

Author: kortirso

CHANGELOG.md

@@ -4,6 +4,10 @@ All notable changes to this project will be documented in this file.
 The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/)
 and this project adheres to [Semantic Versioning](http://semver.org/spec/v2.0.0.html).
 
+## Unreleased
+### Added
+- preprocessing normalization for data set
+
 ## [0.1.4] - 2018-12-17
 ### Modified
 - errors conditions and tests for KNN

lib/learn_kit/knn.ex

@@ -5,7 +5,7 @@ defmodule LearnKit.Knn do
 
   defstruct data_set: []
 
-  alias LearnKit.{Knn}
+  alias LearnKit.Knn
 
   use Knn.Classify
 

lib/learn_kit/knn/classify.ex

@@ -102,17 +102,6 @@ defmodule LearnKit.Knn.Classify do
         end)
       end
 
-      #defp calc_distance_between_points(acc, feature_from_data_set, feature, current_index, size) when current_index <= size do
-      #  Enum.at(feature_from_data_set, current_index) - Enum.at(feature, current_index)
-      #  |> :math.pow(2)
-      #  |> Math.summ(acc)
-      #  |> calc_distance_between_points(feature_from_data_set, feature, current_index + 1, size)
-      #end
-
-      #defp calc_distance_between_points(acc, _, _, _, _) do
-      #  :math.sqrt(acc)
-      #end
-
       defp calc_feature_weight(weight, distance) do
         case weight do
           "uniform" -> 1

lib/learn_kit/preprocessing.ex

@@ -0,0 +1,56 @@
+defmodule LearnKit.Preprocessing do
+  @moduledoc """
+  Module for data preprocessing
+  """
+
+  alias LearnKit.Preprocessing
+
+  use Preprocessing.Normalize
+
+  @type row :: [number]
+  @type matrix :: [row]
+
+  @doc """
+  Normalization of data set
+
+  ## Parameters
+
+    - features: list of features for normalization
+    - options: keyword list with options
+
+  ## Options
+
+    - type: minimax/z_normalization, default is minimax, optional
+
+  ## Examples
+
+      iex> LearnKit.Preprocessing.normalize([[1, 2], [3, 4], [5, 6]])
+      [
+        [0.0, 0.0],
+        [0.5, 0.5],
+        [1.0, 1.0]
+      ]
+
+      iex> LearnKit.Preprocessing.normalize([[1, 2], [3, 4], [5, 6]], [type: "z_normalization"])
+      [
+        [-1.224744871391589, -1.224744871391589],
+        [0.0, 0.0],
+        [1.224744871391589, 1.224744871391589]
+      ]
+
+  """
+  @spec normalize(matrix) :: matrix
+  @spec normalize(matrix, list) :: matrix
+
+  def normalize(features) when is_list(features) do
+    normalize(features, [type: "minimax"])
+  end
+
+  def normalize(features, options) when is_list(features) and is_list(options) do
+    options = Keyword.merge([type: "minimax"], options)
+    case Keyword.get(options, :type) do
+      "z_normalization" -> z_normalization(features)
+      _ -> minimax_normalization(features)
+    end
+  end
+end

lib/learn_kit/preprocessing/normalize.ex

@@ -0,0 +1,56 @@
+defmodule LearnKit.Preprocessing.Normalize do
+  @moduledoc """
+  Module for data normalization
+  """
+
+  alias LearnKit.Math
+
+  defmacro __using__(_opts) do
+    quote do
+      defp minimax_normalization(features) do
+        features_by_index = Math.transpose(features)
+        list_of_params =
+          features_by_index
+          |> Enum.map(fn list ->
+            {
+              Enum.min(list),
+              Enum.max(list)
+            }
+          end)
+        features_by_index
+        |> Enum.with_index()
+        |> Enum.map(fn {feature, index} ->
+          params_for_point = Enum.at(list_of_params, index)
+          range = elem(params_for_point, 1) - elem(params_for_point, 0)
+          feature
+          |> Enum.map(fn point ->
+            (point - elem(params_for_point, 0)) / range
+          end)
+        end)
+        |> Math.transpose()
+      end
+
+      defp z_normalization(features) do
+        features_by_index = Math.transpose(features)
+        list_of_params =
+          features_by_index
+          |> Enum.map(fn list ->
+            {
+              Math.mean(list),
+              Math.standard_deviation(list)
+            }
+          end)
+        features_by_index
+        |> Enum.with_index()
+        |> Enum.map(fn {feature, index} ->
+          params_for_point = Enum.at(list_of_params, index)
+          feature
+          |> Enum.map(fn point ->
+            (point - elem(params_for_point, 0)) / elem(params_for_point, 1)
+          end)
+        end)
+        |> Math.transpose()
+      end
+    end
+  end
+end

test/learn_kit/preprocessing_test.exs

@@ -0,0 +1,33 @@
+defmodule LearnKit.PreprocessingTest do
+  use ExUnit.Case
+
+  alias LearnKit.Preprocessing
+
+  describe "for invalid data" do
+    test "use preprocessor with invalid data" do
+      assert_raise FunctionClauseError, fn ->
+        Preprocessing.normalize("")
+      end
+    end
+
+    test "use preprocessor with invalid options" do
+      assert_raise FunctionClauseError, fn ->
+        Preprocessing.normalize([[1, 2], [3, 4], [5, 6]], "")
+      end
+    end
+  end
+
+  describe "for valid data" do
+    test "normalize data set with minimax normalization" do
+      result = Preprocessing.normalize([[1, 2], [3, 4], [5, 6]])
+
+      assert result == [[0.0, 0.0], [0.5, 0.5], [1.0, 1.0]]
+    end
+
+    test "normalize data set with z normalization" do
+      result = Preprocessing.normalize([[1, 2], [3, 4], [5, 6]], [type: "z_normalization"])
+
+      assert result == [[-1.224744871391589, -1.224744871391589], [0.0, 0.0], [1.224744871391589, 1.224744871391589]]
+    end
+  end
+end