Exploratory Attributes Search for Time-Series Data: An Experimental System for Agricultural Application

摘要

This paper reports an experimental agricultural datamining system whcih purposes to find weather patterns influencing yield of rice. Necessary data for this system are separately maintained in various databases. We then show hos this system integrate them into one database with an assistance of support databases. Next we discuss the attribute selection problem for the data in the integrated database. Our method first exploratory search for a candidate set of attributes. In this case, the support databases is used to avoid a searching space explosion. Once the candidate set is identified, we apply a greedy search in the set of find the most useful subset of attributes.