Architecture of Self-Organizing Map for Time Series Data

Abstract

Self-Organizing Maps (SOM) apply neighborhood learning to enable the creation of close output for similar input. This feature is effective for classification in complex problems. On the other hand, normal SOM cannot directly handle time-series data. For this research we investigate methods applying SOM to classify time-series data. We used delayed-time units in order to map time-series data into the input pattern space.The emission of dioxins from waste incinerators is one of the most important environmental problems today. It is known that optimization of waste incinerator controllers is a very difficult problem due to the complex nature of the dynamic environment within the incinerator. For this experiment, we used sensor data from a waste in cinerator plant as the time-series data, and aimed at classifying instances of dioxin emission. Through computer simulation, we showed that the tested SOM method gave good results in classifying low and high dioxin emission combustion states.