A piecewise linear representation based on compression ratio

As an important method of data preprocessing, some algorithms about piecewise linear representation have been proposed. However, one or more threshold parameters must be input firstly in these algorithms. It is hard to determine these parameters because different time series have different characteristics. So the users have to try and test for many times. Comparing with the existing effective algorithms, the proposed algorithm in this paper only need the compression ratio and then it selects the segmentation points by two iterations. In the first iteration, it selects all the extreme points. Then in the second iteration, it abandons or adds the segmentation points based on the compression ratio and the number of extreme points. The criterion in the second iteration is the vertical distance. At last, two time series are used to verify the effectiveness of the proposed algorithm and the results show that the fitting error of the proposed algorithm is smaller than that of the existing SEEP algorithm and the algorithm based on important points.