Parallel algorithms for mining association rules in time series data
Biplab Kumer Sarker, Takaki Mori, Toshiya Hirata, Kuniai Uehara · 2003
A tremendous growing interest in finding dependency among patterns has been developing in the domain of time series data mining. Researchers have been considering time series data that occurs in the science and business endeavors such as: prices of stocks, the data of a intensive care unit, data flowing from a robot’s sensor, human motion data etc. It is quite effective to find how current and past values in the streams of data are related to the future. However, these data sets with high dimensionality are enormous in size results in possibly large number of mined dependencies. This strongly motivates the need of efficient parallel algorithms. In this paper, we introduce two parallel algorithms to discover dependency from the large amount of motion data. For example, association rule discovered from motion data about walking is “when right hand is up then the left hand and left knee are down”. Since motion data is multi-stream data of 3-D time series and the amount of data is huge and expensive. Consequently, we introduce the method of extracting sequence of symbols as primitive motions from the motion data by using segmentation and clustering processes. The method includes the discretization of the motion data into the symbols of multi-streams. Then the algorithms are implemented on a shared memory multi-processor system. The experimental results show that a significant improvement of performance is achieved using our algorithms. We achieved good speed-up for the algorithms and thus justifies the inevitability of using parallel techniques for mining huge amount of data in the time series domain.