A study of different quality evaluation functions in the cAnt-Miner(PB) classification algorithm

Abstract

Ant colony optimization (ACO) algorithms for classification in general employ a sequential covering strategy to create a list of classification rules. A key component in this strategy is the selection of the rule quality function, since the algorithm aims at creating one rule at a time using an ACObased procedure to search the best rule. Recently, an improved strategy has been proposed in the cAnt-MinerPB algorithm, where an ACO-based procedure is used to create a complete list of rules instead of individual rules. In the cAntMinerPB algorithm, the rule quality function has a smaller role and the search is guided by the quality of a list of rules. This paper sets out to determine the effect of different rule and list quality functions in terms of both predictive accuracy and size of the discovered model in cAnt-MinerPB. The comparative analysis is performed using 12 data sets from the UCI Machine Learning repository and shows that the effect of the rule quality functions in cAnt-MinerPB is different from the results previously presented in the literature.

Enhanced cAntMinerPB Algorithm for Induction of Classification Rules using Ant Colony Approach

“…As has been previously studied [9,10], rule quality functions have different bias and capture different aspects of the rule (e.g., some might favor consistency over coverage).…”

Investigating the impact of various classification quality measures in the predictive accuracy of ABC-Miner

“…Several works aimed to study the effectiveness of these measures, yet in different classification contexts such as classification rule induction [19], [20], [21], [22], which highlighted the importance of rule quality measure chosen to be used to guide the search. We explore the effect of these various classifications quality evaluation measures in guiding the ACO search to construct effective Bayesian network classifiers.…”

“…Sensitivity × Specificity ( Equation 7) -Used in the Ant-Miner [13] and in the cAnt-Miner P B [20] classification rule discovery algorithms . Sensitivity measures the ratio of the count of true positives to the count of all the positive cases, and the specificity measures the ratio of the count of true negatives to the count of all the negative cases.…”

“…In the classification context, it measures the accuracy only with respect to the true positive count, neglecting the true negatives. This measure has been used in [20], [19] as a classification rule quality function, where it was one of the best performing functions.…”

Evaluating the use of different measure functions in the predictive quality of ABC-miner

“…Several works aimed to study the effectiveness of these measures, yet in different classification contexts such as classification rule induction [6,5,3,4], which highlighted the importance of rule quality measure chosen to be used to guide the search. We explore the effect of these different quality evaluation measures in guiding the ACO search to construct effective BN classifiers.…”