An Analytical Evaluation of Matricizing Least-Square-Errors Curve Fitting to Support High Performance Computation on Large Datasets

摘要

The procedure of Least Square-Errors curve fitting is extensively used inmany computer applications for fitting a polynomial curve of a given degree toapproximate a set of data. Although various methodologies exist to carry outcurve fitting on data, most of them have shortcomings with respect toefficiency especially where huge datasets are involved. This paper proposes andanalyzes a matricized approach to the Least Square-Errors curve fitting withthe primary objective of parallelizing the whole algorithm so that highperformance efficiency can be achieved when algorithmic execution takes placeon colossal datasets.