introduce statistics functions

Author: jbytecode

src/main/scala/matrix.scala

@@ -98,7 +98,5 @@ object Matrix:
 
   def appendRowVec(a: Mat, b: Vec): Mat = (a.transpose :+ b).transpose
 
-  def euclideanDistance(a: Vec, b: Vec): Double =
-    math.sqrt(a.zip(b).map((x, y) => (x - y) * (x - y)).sum)
 
-  
+

src/main/scala/statistics.scala

@@ -0,0 +1,28 @@
+package org.expr.mcdm
+
+object Statistics:
+
+    def mean(a: Vec): Double = a.sum / a.length
+
+    def variance(a: Vec) : Double =
+      val m = mean(a)
+      a.map(x => (x - m) * (x - m)).sum / (a.length - 1.0)
+    
+    def correlation(a: Vec, b: Vec): Double =
+      val ma = mean(a)
+      val mb = mean(b)
+      var va = variance(a)
+      var vb = variance(b)
+      (a.zip(b).map((x, y) => (x - ma) * (y - mb)).sum / math.sqrt(va * vb)) / (a.length - 1.0)
+
+    def correlation(a: Mat): Mat = 
+      val n = a.length
+      val m = a(0).length
+      val cor = Matrix.zeros(m, m)
+      for i <- 0 until m do
+        for j <- 0 until m do
+          cor(i)(j) = correlation(Matrix.getcolat(a, i), Matrix.getcolat(a, j))
+      cor  
+
+    def euclideanDistance(a: Vec, b: Vec): Double =
+        math.sqrt(a.zip(b).map((x, y) => (x - y) * (x - y)).sum)

src/main/scala/topsis.scala

@@ -1,10 +1,11 @@
 package org.expr.mcdm
 
-import org.expr.mcdm.MCDMResult;
-import org.expr.mcdm.Direction;
-import org.expr.mcdm.Matrix;
-import org.expr.mcdm.Vec;
-import org.expr.mcdm.Mat;
+import org.expr.mcdm.MCDMResult
+import org.expr.mcdm.Direction
+import org.expr.mcdm.Matrix
+import org.expr.mcdm.Vec
+import org.expr.mcdm.Mat
+import org.expr.mcdm.Statistics
 
 case class TopsisResult(
     val normalizedMatrix: Array[Array[Double]],
@@ -39,12 +40,12 @@ def topsis(
     }
     val distanceToIdeal = 
       Matrix.applyFunctionToRows(weightedNormalizedMatrix, (row: Vec) =>
-        Matrix.euclideanDistance(row, ideal))
+        Statistics.euclideanDistance(row, ideal))
 
 
     val distanceToAntiIdeal =
       Matrix.applyFunctionToRows(weightedNormalizedMatrix, (row: Vec) =>
-        Matrix.euclideanDistance(row, antiIdeal))
+        Statistics.euclideanDistance(row, antiIdeal))
 
     val scores = distanceToAntiIdeal.zip(distanceToIdeal).map {
       case (danti, di) => danti / (danti + di)

src/test/scala/testmatrix.scala

@@ -231,4 +231,5 @@ class TestMatrix extends munit.FunSuite {
     val expected = Array(Array(1.0, 2.0, 3.0), Array(4.0, 5.0, 6.0), Array(7.0, 8.0, 9.0))
     A.assert(Matrix.elementwise_equal(newmat, expected, 1e-6))
   }
+  
 }

src/test/scala/teststatistics.scala

@@ -0,0 +1,44 @@
+import munit.Assertions as A
+
+import scala.math as math
+import org.expr.mcdm.Matrix
+import org.expr.mcdm.Statistics
+
+class TestStatistics extends munit.FunSuite {
+
+  test("Mean of vector"){
+    val a = Array(1.0, 2.0, 3.0, 4.0, 5.0)
+    val mean = Statistics.mean(a)
+    val expected = 3.0
+    A.assertEquals(mean, expected)
+  }
+  test("Variance of vector"){
+    val a = Array(1.0, 2.0, 3.0, 4.0, 5.0)
+    val variance = Statistics.variance(a)
+    val expected = 2.5
+    A.assertEquals(variance, expected)
+  }
+  test("Correlation of vector with itself"){
+    val a = Array(1.0, 2.0, 3.0, 4.0, 5.0)
+    val covariance = Statistics.correlation(a, a)
+    val expected = 1.0
+    A.assertEquals(covariance, expected)
+  }
+  test("Correlation of two different vectors (-1)"){
+    val a = Array(1.0, 2.0, 3.0, 4.0, 5.0)
+    val b = Array(5.0, 4.0, 3.0, 2.0, 1.0)
+    val covariance = Statistics.correlation(a, b)
+    val expected = -1.0
+    A.assertEquals(covariance, expected)
+  }
+  test("Correlation matrix of columns of matrix"){
+    val a = Array(Array(1.0, 2.0, 3.0), Array(4.0, 5.0, 6.0), Array(7.0, 8.0, 90.0))
+    val corr = Statistics.correlation(a)
+    val expected = Array(
+      Array(1.0, 1.0, 0.880812),
+      Array(1.0, 1.0, 0.880812),
+      Array(0.880812, 0.880812, 1.0))
+    A.assert(Matrix.elementwise_equal(corr, expected, 1e-6))
+  }
+  
+}