Cosine Similarity

The Java code measure the similarity between two vectors using cosine similarity formula. The vector's element can be integer or double of Java datatype.

The cosine similarity of two vectors found by a ratio of  dot product of those vectors and their magnitude.

Cosine Similarity

Dot product of 2-D Vector

Dot product of N-D Vector

Vector Magnitude

Cosine Similarity of 2-D vectors

A class Cosine defined two member functions named "similarity" with parameter type difference, in order to support  parameters  type int and double 2-D vectors. 

.

Both class (static) member function similarity can be invoked with two array parameters, which represents the vectors to measure similarity between them. These vectors must have 2 elements.

for example

the vectors  A =[2 3], B=[3 1]

Dot product of two vectors A.B

             A.B = 2x3 + 3x1
                 = 6  + 3
            A.B   = 9

Magnitude of vector A

             |A| =  sqrt ( 2^2 + 3^2 )         
                 =  sqrt ( 4 + 9 )

                 =  sqrt (13)

                 = 3.605

Magnitude of vector B

             |B| =  sqrt ( 3^2 + 1^2 )         
                 =  sqrt ( 9 + 1 )
                 =  sqrt (10)
                 = 3.162
Cosine similarity of vector A and B

                similarity =  9 / ( 3.605 * 3.162)
                            = 9 / 11.4

                            = 0.789

       

import java.util.Arrays;

public class Cosine {

 
public static double similarity(int vec1[],int vec2[]) 
  {
  
    int dop=vec1[0]*vec2[0] +  vec1[1]*vec2[1];    
    double mag1 =Math.sqrt(Math.pow(vec1[0],2)+Math.pow(vec1[1],2));
    double mag2 =Math.sqrt(Math.pow(vec2[0],2)+Math.pow(vec2[1],2));    
    double csim =dop / (mag1 * mag2);
    
    return csim; 
  }
 
 
public static double similarity(double vec1[],double vec2[]) 
  {
  
   double dop=vec1[0]*vec2[0] +  vec1[1]*vec2[1];    
   double mag1=Math.sqrt(Math.pow(vec1[0],2) + Math.pow(vec1[1],2));
   double mag2=Math.sqrt(Math.pow(vec2[0],2) + Math.pow(vec2[1],2));    
   double csim=dop/ (mag1 * mag2);
    
   return csim; 
  }
 
 
 public static void main(String[] args) 
 {
  
      int vector1 [] = {1,2};
      int vector2 [] = {4,5};
      double csim1= Cosine.similarity(vector1, vector2);
            
      System.out.println("\n Cosine  similarity between two 2D vectors");
      System.out.println("integer 2D vectors");
      System.out.println("Vector 1 :" + Arrays.toString(vector1));
      System.out.println("Vector 2 :" + Arrays.toString(vector2));
      System.out.println("similarity value :" + csim1);
      
      
      double vector3 [] = {5.2, 2.6};
      double vector4 [] = {9.8,7.6};
      System.out.println("double 2D vectors");
      double csim2=Cosine.similarity(vector3, vector4);
      System.out.println("Vector 1 :" + Arrays.toString(vector3));
      System.out.println("Vector 2 :" + Arrays.toString(vector4));
      System.out.println("similarity value :" + csim2);
      
                              
 }

}

Output

Cosine similarity between two 2D vectors

integer 2D vectors
Vector 1 :[1, 2]
Vector 2 :[4, 5]
similarity value :0.9778024140774094

double 2D vectors
Vector 1 :[5.2, 2.6]
Vector 2 :[9.8, 7.6]
similarity value :0.9808570865254814

Cosine Similarity for N-Dimensional vectors 

A class Cosine defined two member functions named "similarity" with parameter type difference, in order to support  parameters  type int and double of N dimensional vectors. 

.

Both member function can be invoked with two array parameters, which represents the vectors to measure similarity between them. These vectors must have N elements.

import java.util.Arrays;

public class Cosinend {

 
 public static double similarity(int vec1[],int vec2[]) 
 {
  
    int dop=0;
    for (int n=0;n<vec1.length;n++)
       dop +=vec1[n]*vec2[n] ;
    
     double mag1=0.0,mag2=0.0;    
     for (int n=0;n<vec1.length;n++) 
   {
     mag1 += Math.pow(vec1[n],2) ;
     mag2 += Math.pow(vec2[n],2) ;
   }
   
   mag1=Math.sqrt(mag1);
   mag2=Math.sqrt(mag2);
   
     double csim = dop / (mag1 * mag2);    
     return csim; 
 }
 
 
 public static double similarity(double vec1[],double vec2[]) 
 {  
     double dop=0;
     for (int n=0;n<vec1.length;n++)
        dop +=vec1[n]*vec2[n] ;
     
   double mag1=0.0,mag2=0.0;    
   for (int n=0;n<vec1.length;n++) 
     {
    mag1 +=  Math.pow(vec1[n],2) ;
    mag2 +=  Math.pow(vec2[n],2) ;
      }
    
     mag1=Math.sqrt(mag1);
     mag2=Math.sqrt(mag2);
    
    double csim = dop / (mag1 * mag2);    
    return csim;       
 }
 
public static void main(String[] args) 
  {

         int vector1 [] = {2,2,3,5};
  int vector2 [] = {4,1,2,1};
  double csim1= Cosinend.similarity(vector1, vector2);
             
     System.out.println("\n Cosine  similarity between two N-D vectors");
     System.out.println("integer N-D vectors");
     System.out.println("Vector 1 :" + Arrays.toString(vector1));
     System.out.println("Vector 2 :" + Arrays.toString(vector2));
     System.out.println("similarity value :" + csim1);
       
       
 double vector3 [] = {5.2,2.6,1.4,5.0};
 double vector4 [] = {9.8,7.6,3.2,4.6};
 System.out.println("double ND vectors");
 double csim2=Cosinend.similarity(vector3, vector4);
      System.out.println("Vector 1 :" + Arrays.toString(vector3));
      System.out.println("Vector 2 :" + Arrays.toString(vector4));
      System.out.println("similarity value :" + csim2);
   }

}

Output

Cosine  similarity between two N-D vectors

integer N-D vectors
Vector 1 :[2, 2, 3, 5]
Vector 2 :[4, 1, 2, 1]
similarity value :0.6908492797077574

double N-D vectors
Vector 1 :[5.2, 2.6, 1.4, 5.0]
Vector 2 :[9.8, 7.6, 3.2, 4.6]
similarity value :0.9257252592721228

Related Post
Euclidean distance between two vectors
Manhattan distance between two vectors
Find Chebyshev distance between two points which can be scalars or vectors
Java code measures distance between two points using minkowski distance equation

References
Cosine similarity