Optimal Line Fitting Using Genetic Algorithms

摘要

Genetic algorithms are computational techniques which, given an optimizationproblem, use elements of directed and stochastic search to find the 'best' solution from the space of potential solutions. We apply GA's to the problem of fitting the minimum least-squares piecewise linear function to a set of data points in R(2) . We assume that the number of pieces is known but the knot locations are unknown. The effectiveness of our algorithm is demonstrated with two examples. Results are found to be quite promising and encourage further research.