Link: Least squares

Weighted least squares is a method for Fitting a curve to data. It’s similar to least squares method, but also take weighting into consideration.

The idea is:

  1. Divide the data points in different chunks
  2. In each chunk, use the weighted least squares: the closer the data point is to the focal point (in x axis), the more weighting it has; the last point has less influence on the fitting line.