A set-cover-based approach for the test-cost-sensitive attribute reduction problem
In data mining application, the test-cost-sensitive attribute reduction is an important task which aims to decrease the test cost of data. In operational research, the set cover problem is a typical optimization problem and has a long investigation history compared to the attribute reduction problem. In this paper, we employ the methods of set cover problem to deal with the test-cost-sensitive attribute reduction. First, we equivalently transform the test-cost-sensitive reduction problem into the set cover problem by using a constructive approach. It is shown that computing a reduct of a decision system with minimal test cost is equal to computing an optimal solution of the set cover problem. Then, a set-cover-based heuristic algorithm is introduced to solve the test-cost-sensitive reduction problem. In the end, we conduct several numerical experiments on data sets from UCI machine learning repository. Experimental results indicate that the set-cover-based algorithm has superior performances in most cases, and the algorithm is efficient on data sets with many attributes.
KeywordsAttribute reduction Decision table Rough set Set cover problem Test cost
- Chen CY, Li ZG (2004) A study of reduction of attributes and set covering problem. Comput Eng Appl 2:1–14Google Scholar
- Lavrac N, Gamberger D, Turney P (1996) Cost-sensitive feature reduction applied to a hybrid genetic algorithm. In: Proceedings of the 7th international workshop on algorithmic learning theory, ALTGoogle Scholar
- Quan GR, Hong BR, Ye F, Ren SJ (1998) A heuristic function algorithm for minimum set-covering problem. J Softw 9:156–160Google Scholar
- Slavík P (1996) A tight analysis of the greedy algorithm for set cover. In: Proceedings of the 28th annual ACM symposium on theory of computing, STOC ’96, ACM, pp 435–441Google Scholar