TY - GEN
T1 - SILEA
T2 - 7th International Conference on Information, Intelligence, Systems and Applications, IISA 2016
AU - Aksoy, Ahmet
AU - Gunes, Mehmet Hadi
N1 - Publisher Copyright:
© 2016 IEEE.
PY - 2016/12/14
Y1 - 2016/12/14
N2 - This paper presents SILEA (a System for Inductive LEArning), an efficient inductive learning algorithm for rule extraction. SILEA is a covering algorithm which extracts IF-THEN rules from a collection of examples in a reliable way. The algorithm eliminates exhaustive feature selection by reducing the number of attributes(features) to be considered for each necessary iteration of rule extraction. For every iteration, depending on the number of conditions, it prioritizes numerous attributes over the others to reduce the large number of attribute combinations. This prioritization, however, needs to be done attentively to prevent loss in performance or possibly improve it. SILEA employs the entropy measure for such purpose. As the entropy value decreases for an attribute, its predictability increases. SILEA favors the lower entropy-valued attributes for rule extraction. Another important factor in preserving or improving the performance of the algorithm is the rule extraction and selection procedure. SILEA induces every possible rule for the given combination and selects the most classifying ones among them. It also eliminates rules which might become obsolete by the existence of rules with higher classification performance. In conjunction of these two features, i.e., entropy based attribute prioritization and redundant rule elimination, SILEA extracts rules both accurately and efficiently. The paper describes how the algorithm functions along with its features and discusses its performance compared to some of the well-known algorithms in the field on a number of different data sets.
AB - This paper presents SILEA (a System for Inductive LEArning), an efficient inductive learning algorithm for rule extraction. SILEA is a covering algorithm which extracts IF-THEN rules from a collection of examples in a reliable way. The algorithm eliminates exhaustive feature selection by reducing the number of attributes(features) to be considered for each necessary iteration of rule extraction. For every iteration, depending on the number of conditions, it prioritizes numerous attributes over the others to reduce the large number of attribute combinations. This prioritization, however, needs to be done attentively to prevent loss in performance or possibly improve it. SILEA employs the entropy measure for such purpose. As the entropy value decreases for an attribute, its predictability increases. SILEA favors the lower entropy-valued attributes for rule extraction. Another important factor in preserving or improving the performance of the algorithm is the rule extraction and selection procedure. SILEA induces every possible rule for the given combination and selects the most classifying ones among them. It also eliminates rules which might become obsolete by the existence of rules with higher classification performance. In conjunction of these two features, i.e., entropy based attribute prioritization and redundant rule elimination, SILEA extracts rules both accurately and efficiently. The paper describes how the algorithm functions along with its features and discusses its performance compared to some of the well-known algorithms in the field on a number of different data sets.
UR - http://www.scopus.com/inward/record.url?scp=85013141209&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85013141209&partnerID=8YFLogxK
U2 - 10.1109/IISA.2016.7785430
DO - 10.1109/IISA.2016.7785430
M3 - Conference contribution
AN - SCOPUS:85013141209
T3 - IISA 2016 - 7th International Conference on Information, Intelligence, Systems and Applications
BT - IISA 2016 - 7th International Conference on Information, Intelligence, Systems and Applications
Y2 - 13 July 2016 through 15 July 2016
ER -