Study And Application Of Time Series Similarity And Forecasting Algorithm

Time series analysis is widely used in various fields nowadays. Refer to the research, similarity is the basic problem and forecasting analysis is one of the most important issues. To solve the similarity and forecasting problems in both medical science and passenger flow analysis of urban railway transit, the thesis studies and discusses the following four questions:dynamic analysis and prognosis of boolean time series of stroke symptoms and syndromes; the characteristics analysis of stroke EEG data and prognosis; the lesion location and the characteristics analysis of stroke EEG signals; the similarity analysis and forecasting research of passenger flow data in urban railway transit. The main works and key innovations in the dissertation are summarized as follows:1. The binary time series of stroke symptoms and syndromes are analyzed. From the perspective of Chinese medicine, the characteristics of binary time series of stroke symptoms and syndromes are discussed. An algorithm of mining association rules for binary time series is proposed and the prognosis can be judged by the information of dynamic changes in the binary time series of stroke symptoms and syndromes.2. The EEG time series of stroke is analyzed. From the perspective of western medicine, the characteristics of normal EEG signals and those of anomaly EEG signals with stroke are discussed respectively. A bilateral symmetry similarity index is proposed and it is used to distinguish between the normal EEG and the stroke EEG and also used to judge the prognosis.3. The relationship between stroke lesion location and the quantitative EEG is further analyzed. Several characteristics of stroke EEG are discussed and a partition lesion location method of stroke EEG analysis is proposed.4. The passenger flow time series of urban railwy transit is analyzed and the cyclical characteristics of railway passenger flow time series are discussed. The existing methods of similarity analysis and forecasting in urban road traffic flow are improved and applied to the railway passenger flow time series analysis. A similar pattern based long-term prediction algorithm of railway passenger flow is proposed.