QUESTIONS:

Find the maximum and minimum values for each behavior:

1

The dataset needs to be normalized

3

Calculate similarity between customers.

Ensure that the algorithm inperforms faster

this algorithm is not working for all of the requirements can anyone help?

public class SimilarCustomer {

public int findMostSimilarCustomer(double[][] customerBehaviorMatrix, int currentCustomerIndex) {

//1. find the min max values

double[][] minMaxValuesMatrix = new double[2][customerBehaviorMatrix[0].length];

//assign min and max values to return matrix

for (int j = 0; j < minMaxValuesMatrix[0].length; j++) {

minMaxValuesMatrix[0][j] = Double.MAX_VALUE;

minMaxValuesMatrix[1][j] = Double.MIN_VALUE;

}

//for each column find min and max

for (int i = 0; i < customerBehaviorMatrix.length; i++) {

for (int j = 0; j < customerBehaviorMatrix[i].length; j++) {

//check min

if (customerBehaviorMatrix[i][j] < minMaxValuesMatrix[0][j]) {

//store min in the corresponding return matrix

minMaxValuesMatrix[0][j] = customerBehaviorMatrix[i][j];

}

//check max

if (customerBehaviorMatrix[i][j] > minMaxValuesMatrix[1][j]) {

//store max in the corresponding return matrix

minMaxValuesMatrix[1][j] = customerBehaviorMatrix[i][j];

}

}

}

//2. Normalize the dataset

double[][] normalizedCustomerbehavior =

new double[customerBehaviorMatrix.length][customerBehaviorMatrix[0].length];

// for each entry normalize value and store in return matrix

for (int j = 0; j < normalizedCustomerbehavior[0].length; j++) {

double min = minMaxValuesMatrix[0][j];

double max = minMaxValuesMatrix[1][j];

for (int i = 0; i < normalizedCustomerbehavior.length; i++) {

double value = customerBehaviorMatrix[i][j];

//Assign the normalized value

if ((max-min)==0) {

normalizedCustomerbehavior[i][j] = 0;

} else {

normalizedCustomerbehavior[i][j] = (value - min)/(max-min);

}

}

}

//step 3 and 4 combined. find similar (other) customer; min distance

int returnIndex = -1;

double distance, mindistance = Double.MAX_VALUE;

for (int i = 0; i < normalizedCustomerbehavior.length; i++) {

if (currentCustomerIndex != i) {

//define sum of squares

double sum = 0.0;

for (int j = 0; j < normalizedCustomerbehavior[i].length; j++) {

//calculate difference

double difference = (normalizedCustomerbehavior[currentCustomerIndex][j]-normalizedCustomerbehavior[i][j]);

//square the difference

double squareOfDifference = difference*difference;

//add to the sum

sum += squareOfDifference;

}

//add square root of sum as similarity

distance = Math.sqrt(sum);

// compare distance

if(distance < mindistance)

{

returnIndex = i;

mindistance = distance;

}

}

}

return returnIndex;

}

}