Time Series Piecewise Linear Representation Based on Trend Feature Points

As local maximum and minimum can reflect the trend feature of subsequence, an approach of time series piecewise linear representation based on trend feature point is proposed. The points with large fluctuation can be extracted by judging the variation amplitude of slope. The results of the experiment show that the LMMS algorithm can meet the requirements of different compression ratios, and ensure small fitting error, stable performance, and good adaptability in time series datasets with low volatility. It has nice fitting effect in the volatile datasets under the low compression ratio condition. The relationship between the fitting error of piecewise linear representation algorithm with the whole fluctuation ratio of data and compression ratio is discussed. It can provide a certain reference for time series data mining effectively.