2008 Reports
From Classification Rules to Action Recommendations
Rule induction has attracted a great deal of attention in Machine Learning and Data Mining. However, generating rules is not an end in itself because their applicability is not straightforward especially when the number of rules is large. Ideally, the user would ultimately like to use these rules to decide which actions to take. In the literature, this notion is usually referred to as actionability. The contribution of this paper1 is two-fold: first we propose a survey of the main approaches developed to address actionability. This topic has received growing attention in the past years. We present a classification of the main research in this area as well as a comparative study between the different approaches. Second, we propose a new framework to address actionability. Our goal is to lighten the burden of analyzing a large set of classification rules when the user is confronted with an "unsatisfactory situation" and needs help to decide what appropriate actions to take in order to remedy the situation. The method consists in comparing the situation to a set of classification rules. This is achieved by using a suitable distance that allows one to suggest action recommendations requiring minimal changes to improve the situation. We propose the algorithm DAKAR for learning action recommendations and we present an application to environment protection. Our experiment shows the usefulness of our contribution for action recommendation but also raises some concerns about the impact of the redundancy of a set of rules in learning action recommendations of good quality.
Subjects
Files
- CCLS-08-01.pdf application/pdf 667 KB Download File
More About This Work
- Academic Units
- Center for Computational Learning Systems
- Publisher
- Center for Computational Learning Systems, Columbia University
- Series
- CCLS Technical Report, CCLS-08-01
- Published Here
- November 12, 2010