New technique reduces over-fitting in polynomial regression, improving accuracy.
A new technique called PALS was created to reduce over-fitting in polynomial regression. PALS minimizes the areas between the polynomial curve and line segments connecting data points. It uses a simple system of equations to find the best polynomial fit. Testing showed that PALS can create polynomials with less over-fitting, even with high polynomial orders. PALS doesn't need any special parameters for fitting, but it only works for polynomials with one independent variable.