Minimising the delta test for variable selection in regression problems

摘要: The problem of selecting an adequate set variables from a given data sampled function becomes crucial by the time designing model that will approximate it. Several approaches have been presented in literature although recent studies showed how delta test is powerful tool to determine if subset correct. This paper presents new methodologies based on such as tabu search, genetic algorithms and hybridisation them, which representative function. considers well scaling where relevance value assigned each variable. were adapted be run parallel architectures so better performances could obtained small amount time, presenting great robustness scalability.