Design of Query-Driven System for Time-Utility Based Data Mining on Medical Data

Abstract

Association rule mining(ARM) techniques search for groups of frequently co-occurring items (i.e., frequent itemset) in a market-basket transaction database and convert these groups into business-oriented rules. The problem of ARM will gain momentum when it is attached with the time of transaction. High utility itemset mining is a research area of utility based data mining, aimed at finding itemsets that contribute most to the total utility. The association of time and utility on frequent itemsets gives a novel approach to efficiently capture the transactions for getting better predictions and planning for an enterprise. Previous research has focused mainly on how to obtain exhaustive lists of association rules. However, users often prefer a quick response to targeted queries. To accelerate the processing of such queries, a query-driven system called TD-FVAUFM (Time-Dependent Fast Value Added Utility Frequent Mining) is proposed in this paper. It performs data preprocessing steps on the given database and the resultant database is converted in the form of an itemset tree, a compact data structure suitable for query response. The proposed system is applied on a medical database containing patient’s records. It generates association rules that predict possible diseases with risk factor and frequency with respect to time. Experiments indicate that the targeted queries are answered in a time that is roughly linear in the number of transactions.