Proposed Design for Data Retrieval using Efficient Algorithm

Abstract

Data mining is used for finding patterns from large amount of data which is in raw format. These patterns are then analyzed to gain useful information from them. There are many branches of data mining, one of the most interesting branch is frequent item-set mining (FIM). FIM deals with finding items that are frequently brought together by customers. Like for example, if a customer purchases a mobile phone, he also tends to purchase mobile cover, ear phones etc along with it. But such kinds of patterns are not always useful to all stake-holders. Such patterns do not emphasize on the profit obtained of sale i.e. the utility obtained from product. In order to overcome this problem, the concept of high utility item-set mining (HUIM) came into existence. HUIM is used to find the utility or profit obtained from the items in transaction data. There are various algorithms for HUIM, TKU (Top K Utility) and TKO (Top K in one phase) are two well known algorithms of HUIM. The detailed study and practical analysis of these two algorithms show that there are certain drawbacks assigned with them. TKO algorithm gets executed in very less amount of time but it gives incorrect output. Whereas TKU algorithm gives accurate results when applied on database, but its execution time is very high. Hence in order to enhance the performance of these two HUIM algorithms a hybrid algorithm i.e. TKO with TKU algorithm is proposed in this paper. The two algorithms when combined give accurate result and also get executed in considerable less amount of time