Discovery Association Rules in Time Series Data

Abstract

Rule discovery from time series data is a data mining technique that tries to find relationshipsof sequential data. Finding association rules from time series data is different from finding suchrules in traditional data because time series data is orderly data with a sequence that must bepreserved. Many researchers have proposed many methods of analyzing and mining time seriesdata, but most of them did not focus on finding association rules, and the data used in theirexperimentations were discretized symbols. In fact, many situations collect data in continuousnumeration time series. In this paper, we propose a novel technique to find association rules fromtime series data. Our technique can analyze either the numerical time series or the symbolic timeseries and show the resulting rules as X --> Y , which means that the group of pattern Y shouldoccur within time t when the group of pattern X occurs.

Keywords : Time Series / Association Rules / Sequential Data / Data Mining