In this work it is presented a methodological proposal to build models for Time Series with missing and erroneous values. This methodology consist of two stages: first, it is realized an estimating of the missing and erroneous values of the series; and second, it is built a model for the series. The proposal is based on Self Adaptive Genetic Algorithms that were especially utilized to calculate ARMA models for the NN5-REDUCED problems which results are presented in this work. This methodology here presented can be generalized for the treatment of this type of Time Series by other non linear models that use, for example, neuronal networks, fuzzy logic, etc.
|Translated title of the contribution||Conferencia Internacional en Inteligencia Artificial. Genética y Métidos Evolucionarios (GEM): Modelo de Series de Tiempo para detección de valores incorrectos utilizando Algoritmos Genéticos.|
|Original language||American English|
|Title of host publication||International Conference Inteligence Artificial. General Topics in Artificial Intelligence. Genetic and Evolutionary Methods (GEM)|
|Subtitle of host publication||Modeling Time Series with Missing and Incorrect Values Using Self Adaptive Genetic Algorithms|
|Place of Publication||Estados Unidos de Norteamerica|
|Number of pages||6|
|State||Published - 15 Jul 2011|