Copyright © 2018 Authors. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. International Journal of Engineering & Technology, 7 (4.36) (2018) 533-541 International Journal of Engineering & Technology Website: www.sciencepubco.com/index.php/IJET Research paper Finding Efficient Positive and Negative Itemsets Using Interestingness Measures P. Asha 1* , T. Prem Jacob 2 , A. Pravin 3 1 Asst.Prof, Department of Computer Science and Engineering, Sathyabama Insttitue of Science and Technology, Chennai. 2 Asst.Prof, Department of Computer Science and Engineering, Sathyabama Insttitue of Science and Technology, Chennai. 3 Asst.Prof, Department of Computer Science and Engineering, Sathyabama Insttitue of Science and Technology, Chennai. *Corresponding author E-mail:ashapandian225@gmail.com Abstract Currently, data gathering techniques have increased through which unstructured data creeps in, along with well defined data formats. Mining these data and bringing out useful patterns seems difficult. Various data mining algorithms were put forth for this purpose. The associated patterns generated by the association rule mining algorithms are large in number. Every ARM focuses on positive rule mining and very few literature has focussed on rare_itemsets_mining. The work aims at retrieving the rare itemsets that are of most interest to the user by utilizing various interestingness measures. Both positive and negative itemset mining would be focused in this work. Keywords: Association rule, positive association rules, negative association ruls, Interestingness measures. 1. Introduction Data Mining has grasped people’s interest due to the availability of a wide range of raw data where useful information is poor. Thus, there is a need for extracting meaningful information from the large amount of data that is of user understandable form. Data Mining is an analysis process where patterns are mined from a large database or repositories. To retrieve useful information, data preprocessing techniques such as Data cleaning and integrating, after selection of data followed by transformation using which the mining is done and such data representation is popularly known as Knowledge Discovery in KDD. Mining is performed on relational databases, Data Warehouse, Transaction Databases and Advanced databases such as OODB, ORDB, Multimedia, Spatial and Web Mining Databases. Mining can be applied in various fields such as Sales/Marketing, Healthcare/Insurance, Banking/Finance, Medicine, Biomedicine, Transportation, Telecommunication etc.It is advantageous as a powerful tool that find patterns and relationships among data which helps in discovering hidden information from the large and useless datasets available. But Data Mining cannot work without human effort and also it cannot tell the value of information mined for our need. Thus to ensure meaningful Data Mining results, the researcher must understand the data available. Association Rule can otherwise be called as pattern. It has two constituents: antecedent and consequent which is similar to if - then respectively. The item P corresponds to the antecedent endowed in the data whereas the consequent Q is endowed in the combination of the antecedent P. Patterns that are mined must be meaningful. Such a meaningful information or pattern is discovered through some interesting measures. Most important measure is supported that indicates how frequently the itemset P or Q occur in the dataset and confidence attests the number of counts, the if/then (P->Q) statements have been committed to be true. Other measures are Laplace, Pearson coefficient, conviction, P-S measure, interest factor, chi-square test, lift, leverage. Some of these measures are discussed and efficient method is comparatively determined in the study. Also the relationship between these meaningful pattern can be identified through a traditional method called Association Rule Mining (ARM). ARM is becoming a research topic that mines associated rules. The strength of the associated rules is determined by interestingness measure. Mined rules have to satisfy some user specified minimum value of support and the confidence. Pretty good algorithms such as Apriori, Elcat, FP-growth are available to generate association rules. In mining, association rules favours us to analyze and predict the customer’s behavior. Also, it plays an starring role in the market basket analysis and many other real time applications. ARM finds relationship between two itemsets of the pattern, P→Q, in which P and Q are disjoint items. Likewise ARM can be used to mine k-itemsets. Most of the existing research is on mining positive_association_rules of the pattern P→Q, but we also focus on mining negative association rules of the pattern, P→~Q, ~P→Q, ~P→~Q. This absence of itemsets is also considered in our study. Different methods of ARM are discussed which mines frequent and infrequent itemsets from where both positive and negative rules are mined. Strong rules can be mined by applying interestingness measure. Major usage of mining negative association rules are fraud detection and to find genetic disorders in the field of Bioinformatics. Primary issues in mining negative rules are identification of appropriate patterns from the large database, thus making it a challenging attempt. Also, various methods that were incorporated to lessen the number of rules generated and number of scans to the database is the main objective discussed in the study. 2. Review on existing work Rakesh et al. compared Apriori and Apriori_tid algorithms for Association Rule Mining and combined the advantages of both the algorithms, which was termed asApriori_hybrid. Apriori