Multiple-model polynomial regression and efficient algorithms for data analysis

Abstract

This paper newly proposes a data analysis method using multiple-model p-order polynomial regression (MMPR), which separates given datasets into subsets and constructs respective polynomial regression models for them. An approximate algorithm to construct MMPR models based on $(ϵ,\delta )$-estimator, and mathematical proofs of the correctness and efficiency of the algorithm are introduced. This paper empirically implements the method on both synthetic and real-world datasets, and it's shown to have comparable performance to existing regression methods in many cases, while it takes almost the shortest time to provide a regression model with high prediction accuracy.