48 research outputs found
A Fuzzy Algorithm for Mining High Utility Rare Itemsets -FHURI
No full textClassical frequent itemset mining identifies frequent itemsets in transaction databases using only frequency of item occurrences, without considering utility of items. In many real world situations, utility of itemsets are based upon user's perspective such as cost, profit or revenue and are of significant importance. Utility mining considers using utility factors in data mining tasks. Utility-based descriptive data mining aims at discovering itemsets with high total utility is termed High Utility Itemset mining. High Utility itemsets may contain frequent as well as rare itemsets. Classical utility mining only considers items and their utilities as discrete values. In real world applications, such utilities can be described by fuzzy sets. Thus itemset utility mining with fuzzy modeling allows item utility values to be fuzzy and dynamic over time. In this paper, an algorithm, FHURI (Fuzzy High Utility Rare Itemset Mining) is presented to efficiently and effectively mine very-high (and high) utility rare itemsets from databases, by fuzzification of utility values. FHURI can effectively extract fuzzy high utility rare itemsets by integrating fuzzy logic with high utility rare itemset mining. FHURI algorithm may have practical meaning to real-world marketing strategies. The results are shown using synthetic datasets
Business information query expansion through semantic network
No full textIn this article, we propose a method for business information query expansions. In our approach, hypernym/hyponymy and synonym relations in WordNet are used as the basic expansion rules. Then we use WordNet Lexical Chains and WordNet semantic similarity to assign terms in the same query into different groups with respect to their semantic similarities. For each group, we expand the highest terms in the WordNet hierarchies with hypernym and synonym, the lowest terms with hyponym and synonym and all other terms with only synonym. In this way, the contradictory caused by full expansion can be well controlled. Furthermore, we use collection-related term semantic network to further improve the expansion performance. And our experiment reveals that our solution for query expansion can improve the query performance dramatically
A hybrid heuristic approach for attribute-oriented mining
No full textWe present a hybrid heuristic algorithm, clusterAOI, that generates a more interesting generalised table than obtained via attribute-oriented induction (AOI). AOI tends to overgeneralise as it uses a fixed global static threshold to cluster and generalise attributes irrespective of their features, and does not evaluate intermediate interestingness. In contrast, clusterAOI uses attribute features to dynamically recalculate new attribute thresholds and applies heuristics to evaluate cluster quality and intermediate interestingness. Experimental results show improved interestingness, better output pattern distribution and expressiveness, and improved runtime. © 2013 Elsevier B.V
An algorithm to mine general association rules from tabular data
No full textMost methods for mining association rules from tabular data mine simple rules which only use the equality operator "=" in their items. For quantitative attributes, approaches tend to discretize domain values by partitioning them into intervals. Limiting the operator only to "=" results in many interesting frequent patterns that may not be identified. It is obvious that where there is an order between objects, operators such as greater than or less than a given value are as important as the equality operator. This motivates us to extend association rules, from the simple equality operator, to a more general set of operators. We address the problem of mining general association rules in tabular data where rules can have all operators {≤, >, ≠, =} in their antecedent part. The proposed algorithm, mining general rules (MGR), is applicable to datasets with discrete-ordered attributes and on quantitative discretized attributes. The proposed algorithm stores candidate general itemsets in a tree structure in such a way that supports of complex itemsets can be recursively computed from supports of simpler itemsets. The algorithm is shown to have benefits in terms of time complexity, memory management and has good potential for parallelization. © 2009 Elsevier Inc. All rights reserved
Finding associations in composite data sets
No full textIn this paper, a composite fuzzy association rule mining mechanism CFARM, directed at identifying patterns in datasets comprised of composite attributes, is described. Composite attributes are defined as attributes that can take simultaneously two or more values that subscribe to a common schema. The objective is to generate fuzzy association rules using "properties" associated with these composite attributes. The exemplar application is the analysis of the nutrients contained in items found in grocery data sets. The paper commences with a review of the back ground and related work, and a formal definition of the CFARM concepts. The CFARM algorithm is then fully described and evaluated using both real and synthetic data sets
