Study of Knowledge Acquisition Using Rough Set Merging Rule from Time Series Data

Abstract

Rough Set Theory proposed in 1982 by Zdzislaw Pawlak. This theory can be data mining based on decision rules from a database, a web page, a big data, and so on. The decision rule is employed for data analysis as well as calculating an unknown object. We used rough set to analyze time-series data. We obtained prediction knowledge from time series data using decision rules. Economic time-series data was predicted using decision rules. However, when acquiring a decision rule from time series data, there are cases where the number of decision rules is very large. If the number of decision rules is very large, it is difficult to acquire knowledge. We proposed a method of merging them to reduce the number of decision rules. Similar to how it is difficult to acquire knowledge from multiple rules, it is also difficult to acquire knowledge from rules with a large number of condition attributes. Our method reduces the number of conditions attributes and thereby reduces the number of rules. However, it is not always possible to reduce rules. There are cases where the number of rules increases. In this thesis, we examine under what conditions rule reduction is possible. Change the condition attribute and verify the effect on rule reduction. We acquire knowledge using the Nikkei Stock Average. We acquire decision rule by rough set method and consider the influence on rule reduction.