Article Preview
Top1. Introduction
The traditional ARM is a data mining technique, which is aimed to retrieve highly related objects in a database based on its occurrences. This does not take semantic significance of objects into account. Recently utility based association rules were attracting researchers, which consider significant factors of the objects in a database. Preliminaries in utility mining calculations can be found in ((Kumari, Sanjeevi&Rao, (2019); (Zhang, Fang, Sun& Wang (2019); Singh (2019); (Nguyen, Nguyen, Nguyen, Vo, Fournier-Viger, &Tseng (2019)). Generation of utility association rules contain two phases. The first phase is the generation of high utility itemsets and the second phase is the discovery of utility association rules from the resulted first phase utility itemsets. Utility association rules are represented as, I1→ I2, only if, I1 and I2 are high utility itemsets and I1 ∩ I2 is NULL. Since the rules are generated from high utility itemsets, the generated rule is guaranteed to be high utility rule. In traditional association rule mining, support and confidence measures are used to discover strong rules. In utility based association rules, these support and confidence measures are not sufficient. The utility factors for the rules have to be incorporated in rules. In this work, a novel measurement for utility confidence is given as,
Util.conf(I
1→ I
2) = .
(1)So far in the literature, no optimization algorithm was proposed in Utility based Association rule mining. This research work was aimed to retrieve High Utility Association Rules (HUAR) from a sales transaction database, using Cockroach Swarm Optimization algorithm (CSO). The proposed algorithm is named as CSOUAR (Cockroach Swarm Optimization for High Utility Association Rule Mining) algorithm. The methodology is shown in figure 1.
Figure 1. Methodology - Single level High Utility Association Rule Mining
Cockroach Swarm Optimization (CSO) proposed by Cheng & Tang (2010) is one of the recent developments in Swarm Intelligence. Cockroaches belongs to InsectaBlattode and its habits are chasing, swarming, dispersing, being ruthless and omnivorous, and food searching (Xing &Gao, 2014); (Zhang, Agarwal, Bhatnagar, Balochian& Zhang, 2014); (Neogi, 2015). Cockroach can feel the immediate change of surroundings, with the help of its sensitive antennae. Cockroaches interact with peers and respond the surrounding and make decisions based on their interaction such as selecting shelter, searching for food sources and friends, dispersing when danger is noticed, and eating one another when there is food scarce (Cheng, Wang, Song, Guo, 2011). The model of CSO algorithm is grounded on three foremost deeds of cockroaches namely, Chase-Swarming Behaviour, Dispersing Behaviour and Ruthless Behaviour. The chase swarming habit is inspired by the means of communication of cockroaches with each other. The chase-swarming behaviour of cockroach individuals is implemented to search the global optimum of solutions. The dispersing habit is modelled from disperse of cockroaches to the surroundings, while the environment has any sudden change. The dispersing behaviour is used to keep individuals diversity, hence employed for local optimum solution. The ruthless habit is exhibited, when the bigger cockroaches eat the smaller, the stronger eat the weaker during food shortage. The ruthless behaviour can improve the solutions obtained. The basic terminologies used in CSO algorithm was given in Table 1.