Linear-Regression-Simulation

Linear Regression Simulation

Linear regression is a linear approach to modelling the relationship between a scalar response (or dependent variable) and one or more explanatory variables (or independent variables). There are two methods used in the simulation.

1.Least Square Method
2.Gradient descent

#Least Sqaure Method:

void leastSqure() throws Exception{

  float xmean = 0, ymean = 0;
  for(PVector point: points){
    xmean += point.x/points.size();
    ymean += point.y/points.size();
  }

  float numerator = 0,denominator = 0;
  for(PVector point : points){
    numerator += (point.x-xmean)*(point.y-ymean);
    denominator += (point.x-xmean)*(point.x-xmean);
  }
  m = numerator/denominator;    // m slope of the line
  b = ymean - (m*xmean);        // y intercept of the line
}

#Gradient Descent:

void gradientDescent() throws Exception{
  float learning_rate = 0.01;

  for(PVector point : points){
    float x = point.x;
    float y = point.y;
    float guess = (m*x) + b;
    float error = y - guess;
    m = m + (error*x) * learning_rate;    // m is the slope of the line
    b = b + error * learning_rate;        // y intercept of the line
  }
}

download processing at: https://processing.org/