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

The procedure of Least Square-Errors curve fitting is extensively used in many computer applications for fitting a polynomial curve of a given degree to approximate a set of data. Although various methodologies exist to carry out curve fitting on data, most of them have shortcomings with respect to efficiency especially where huge datasets are involved. This paper proposes and analyzes a matricized approach to the Least Square-Errors curve fitting with the primary objective of parallelizing the whole algorithm so that high performance efficiency can be achieved when algorithmic execution takes place on colossal datasets.