Abstract | ||
---|---|---|
Feature or attribute selection is a crucial activity when knowledge discovery is applied to very large databases. Its main objective is to eliminate irrelevant or redundant attributes to obtain a computationally tractable problem, without affecting the classification quality. In this article a novel optimization approach is evaluated. This method uses concave programming to minimize the number of attributes to input to the mining algorithm and also, to minimize the classification error. This technique is evaluated using a billing data base from the national electric utility in Mexico. The results are compared against those obtained by traditional techniques. From this experimentation, several improvements to the optimization approach are suggested. |
Year | DOI | Venue |
---|---|---|
2005 | 10.1007/11554028_9 | KES (4) |
Keywords | Field | DocType |
feature selection,billing data base,attribute selection,electric billing database,knowledge discovery,optimization approach,classification error,computationally tractable problem,crucial activity,classification quality,concave programming,novel optimization approach,very large database | Data mining,Tariffication,Feature selection,Electric utility,Computer science,Concave programming,Redundancy (engineering),Artificial intelligence,Knowledge extraction,Data mining algorithm,Machine learning,Database | Conference |
Volume | ISSN | ISBN |
3684 | 0302-9743 | 3-540-28897-X |
Citations | PageRank | References |
0 | 0.34 | 10 |
Authors | ||
3 |
Name | Order | Citations | PageRank |
---|---|---|---|
Manuel Mejía-Lavalle | 1 | 14 | 7.86 |
Guillermo Rodríguez | 2 | 34 | 6.46 |
Gustavo Arroyo | 3 | 13 | 3.46 |