Computing Reviews
Today's Issue Hot Topics Search Browse Recommended My Account Log In
Review Help
Search
ETARM: an efficient top-k association rule mining algorithm
Nguyen L., Vo B., Nguyen L., Fournier-Viger P., Selamat A. Applied Intelligence48 (5):1148-1160,2018.Type:Article
Date Reviewed: Aug 6 2018

Association rule mining is a central component of data mining that deals with finding associations between elements in a database based on their use. The ubiquitous presence of association rules is sometimes credited to Amazon for their novel application in the e-commerce industry, that is, suggesting other items based on purchase transactions. Nowadays, virtually every content or e-commerce website includes a recommendation engine that suggests other items for a consumer to consume (buy/rent/read/review) based on their history. Thus, the commercial value of association rules is well established. Due to the limited amount of screen real estate available, we are generally only interested in showing the top few recommendations. Furthermore, due to the online nature of e-commerce, association rules need to be continuously updated as more transaction data is captured. To prevent a proliferation of rules, we are generally only interested in the strongest rules, that is, those that keep the recommendation system running efficiently and still recommend items that have high association value.

This interesting paper attempts to find top-k association rules, and do so efficiently. The efficiency of the system is an important metric because most applications in this domain produce and use a very large amount of data. Therefore, a more efficient algorithm can produce rules in a more time- and space-efficient manner, enabling the resulting recommendation system to run more efficiently.

A variety of association rule mining algorithms are available in the literature, and the proposed ETARM algorithm builds upon Fournier-Viger and Tseng’s TopKRules algorithm [1]. ETARM makes two key improvements to TopKRules: (1) “if the largest item in the antecedent (consequent) of a rule according to the lexicographical order is also the largest item in the database, the rule antecedent should not be expanded by a left (right) expansion”; and (2) “if the confidence of a rule is less than [the minimum confidence threshold], it should not be expanded by right expansion, as the resulting rule will not be a top-k rule.”

To show ETARM’s strength and significance, the authors compare the algorithms on six different datasets, including a retail dataset with more than 88000 transactions and more than 16000 items. According to the empirical results presented, ETARM outperforms TopKRules in terms of both runtime and memory utilization, in many cases seeing up to a 50 percent reduction both in runtime and in space utilization.

It may be reasonable to question the paper’s verbose nature. Perhaps it is targeted at beginners who can benefit from a gentle introduction to the concepts. Experienced researchers and practitioners could use a shorter version that simply highlights the key improvements and empirical results.

Reviewer:  Amrinder Arora Review #: CR146189 (1811-0592)
1) Fournier-Viger, P.; Tseng, V. S. Mining Top-k Sequential Rules. In: Advanced data mining and applications (LNAI 7121). 180-194, Springer, 2011.
Bookmark and Share
 
Learning (I.2.6 )
 
Would you recommend this review?
yes
no
Other reviews under "Learning": Date
Learning in parallel networks: simulating learning in a probabilistic system
Hinton G. (ed) BYTE 10(4): 265-273, 1985. Type: Article
Nov 1 1985
Macro-operators: a weak method for learning
Korf R. Artificial Intelligence 26(1): 35-77, 1985. Type: Article
Feb 1 1986
Inferring (mal) rules from pupils’ protocols
Sleeman D.  Progress in artificial intelligence (, Orsay, France,391985. Type: Proceedings
Dec 1 1985
more...

E-Mail This Printer-Friendly
Send Your Comments
Contact Us
Reproduction in whole or in part without permission is prohibited.   Copyright 1999-2024 ThinkLoud®
Terms of Use
| Privacy Policy