Piecewise linear representations of time series using improved bottom-up algorithm

This paper proposes a new algorithm,named as the improved bottom-up algorithm,to approximate time series in a linear manner.The algorithm starts by the finest possible approximation of time series,so that n-1 segments are used to approximate the n-length time series.The fitting errors of merging each pair of adjacent segments are calculated.The algorithm iteratively merges the minimal fitting error pair until the number of segments meets a given value K.The main novelty of the proposed algorithm is to devise a decision function to determine an optimal choice of K.The decision function is composed of two parts,namely,the weighted fitness and trend similarity between the time series and its corresponding approximation.The effectiveness of the proposed algorithm is illustrated via numerical and industrial examples.This paper proposes a new algorithm,named as the improved bottom-up algorithm,to approximate time series in a linear manner.The algorithm starts by the finest possible approximation of time series,so that n-1 segments are used to approximate the n-length time series.The fitting errors of merging each pair of adjacent segments are calculated.The algorithm iteratively merges the minimal fitting error pair until the number of segments meets a given value K.The main novelty of the proposed algorithm is to devi中国自动化学会过程控制专业委员会