Search

Se indica cómo manejar una gran base de datos consistente de series temporales no lineales, aplicando distintas técnicas de modelamiento no lineal a estas series. Aunque no existen guías explícitas de manipulación de series temporales no lineales en la profusa bibliografía actual, existen diferentes enfoques que pueden ser tomados en cuenta. Para ello se estudió una base de datos mensual correspondiente a datos del Fenómeno del Niño (ENSO), entre los años 1866 y 2006. Se explica cómo debe manipularse esta base de datos que poseen características de no linealidad, la cual será usada para hacer pronósticos varios pasos en adelante. Se aplicaron dos test estándar: Información Mutua Promedio (AMI) y Falsos Vecinos más Cercanos (FNN). Se obtuvo el espaciamiento óptimo de los datos, así como el número de datos hacia atrás necesarios para pronosticar valores hacia el futuro. Luego, se diseñaron varios modelos de redes neuronales artificiales (RNA), con diferentes reglas de aprendizajes, funciones de transferencia, elementos de procesamiento (o neuronas) en la capa escondida, etc., que permitieron hacer pronóstico de hasta 20 pasos en adelante. Las mejores redes correspondieron a aquellas que poseían como regla de aprendizaje la Regla Delta y la Regla Extendida, con función de transferencia sigmoide y tangente hiperbólica. El tipo de RNA usada fue una de multicapas alimentada hacia adelante y entrenada mediante la técnica de propagación hacia atrás. Se probaron redes con una, dos capas ocultas y sin ninguna capa. El mejor modelo que se obtuvo resultó ser uno consistente de una capa oculta.We indicate how to handle a large database consisting of nonlinear time series, applying different nonlinear modelling techniques to this kind of times series. Nowadays in the current references there is no explicit guide of how to manipulate data from nonlinear time series; however, there are approaches that can be taken account. To this end we studied a monthly database corresponding to South Oscillation Index (SOI) and between the years 1886 to 2006. It explains how there must manipulated this database whose data possess nonlinear characteristic, which will be used to do forecasts several steps ahead. Two standard tests to this database were applied, the Average Mutual Information (AMI) and the False Nearest Neighbours (FNN). The optimal spacing of the information was obtained as well as the number of values backward necessary to predict values towards the future. Then, several models were designed of artificial neural nets (ANN), with different learning rules, function of transfer, elements of process (or neurons) in the hidden layer, etc., that allowed to do forecasting of up to 20 steps ahead. The best networks were those that possessed the rules of learning called extDBD and Delta-Rule, and sigmoid as well as hyperbolic tangent as function of transfer. The type of used network was one of feedforward multilayer perceptron and trained by means of backpropagation technique. Networks were proved by one, two hidden layers and without any hidden layer. The best model that was obtained it turned out to be one that consisted with an alone hidden layer.

Se diseñó una red neuronal artificial (RNA) para hacer predicciones de valores de concentraciones horarias de material particulado fino en la atmósfera. El estudio está basado en los datos de tres años de series de tiempo de pm2.5 (material particulado suspendido de 2,5 micrones de diámetro), obtenidos en una estación céntrica de la Red MACAM de la ciudad de Santiago de Chile, entre los años 1994 y 1996. Para obtener el espaciamiento óptimo de los datos, así como el número de datos hacia atrás necesarios para pronosticar el valor futuro, se aplicaron dos test estándar usados en estudio de sistemas dinámicos, como Información Mutua promedio (AMI) y Falsos Vecinos más Cercanos (FNN). De esta manera se encontró que lo más conveniente era considerar como entrada los datos de PM2.5 cada seis horas durante un día (cuatro datos), y en base a ellos predecir el dato siguiente. Una vez fijo el número de variables de entrada y elegida la variable a pronosticar, se diseñó un modelo predictivo basado en la técnica de RNA. El tipo de modelo de RNA usado fue uno de multicapas, alimentado hacia adelante y entrenado mediante la técnica de propagación hacia atrás. Se probaron redes sin capa oculta y con una y dos capas ocultas. El mejor modelo resultó ser con una capa oculta, a diferencia de lo obtenido en trabajo anterior que reportaba que la red sin capa oculta era más eficiente. Los resultados fueron más precisos que los obtenidos con un modelo de persistencia (el valor en seis horas más será el mismo que el actual).An artificial neural network for the forecasting of concentrations of fine particulate matter in the atmosphere was designed. The data set analyzed corresponds to three years of pm2.5 time series (particulate matter in suspension with aerodynamic diameter less than 2,5 microns), measured in a station that belongs to Santiago's monitoring network (Red MACAM) and is located near downtown. We consider measurements of concentrations between May and August for years between 1994 and 1996. In order to find the optimal time spacing between data and the number of values into the past necessary to forecast a future value, two standard tests were performed, Average Mutual Information (AMI) and False Nearest Neighbours (FNN). The results of these tests suggest that the most convenient choice for modelling was to use 4 data with 6 hour spacing on a given day as input in order to forecast the value at 6 AM on the following day. Once the number and type of input and output variables are fixed, we implemented a forecasting model based on the neural network technique. We used a feedforward multilayer neural network and we trained it with the backpropagation algorithm. We tested networks with none, one and two hidden layers. The best model was one with one hidden layer, in contradiction with a previous study that found that minimum error was obtained with a net without hidden layer. Forecasts with the neural network are more accurate than those produced with a persistence model (the value six hours ahead is the same as the actual value).

We indicate how to handle a large database consisting of nonlinear time series, applying different nonlinear modelling techniques to this kind of times series. Nowadays in the current references there is no explicit guide of how to manipulate data from nonlinear time series; however, there are approaches that can be taken account. To this end we studied a monthly database corresponding to South Oscillation Index (SOI) and between the years 1866 to 2006. It explains how there must manipulated this database whose data possess nonlinear characteristic, which will be used to do forecasts several steps ahead. Two standard tests to this database were applied, the Average Mutual Information (AMI) and the False Nearest Neighbours (FNN). The optimal spacing of the information was obtained as well as the number of values backward necessary to predict values towards the future. Then, several models were designed of artificial neural nets (ANN), with different learning rules, function of transfer, elements of process (or neurons) in the hidden layer, etc., that allowed to do forecasting of up to 20 steps ahead. The best networks were those that possessed the rules of learning called extDBD and Delta-Rule, and sigmoid as well as hyperbolic tangent as function of transfer. The type of used network was one of feedforward multilayer perceptron and trained by means of backpropagation technique. Networks were proved by one, two hidden layers and without any hidden layer. The best model that was obtained it turned out to be one that consisted with an alone hidden layer. [ABSTRACT FROM AUTHOR]

An artificial neural network for the forecasting of concentrations of fine particulate matter in the atmosphere was designed. The data set analyzed corresponds to three years of pm2.5 time series (particulate matter in suspension with aerodynamic diameter less than 2, 5 microns), measured in a station that belongs to Santiago's monitoring network (Red MACAM) and is located near downtown. We consider measurements of concentrations between May and August for years between 1994 and 1996. In order to find the optimal time spacing between data and the number of values into the past necessary to forecast a future value, two standard tests were performed, Average Mutual Information (AMI) and False Nearest Neighbours (FNN). The results of these tests suggest that the most convenient choice for modelling was to use 4 data with 6 hour spacing on a given day as input in order to forecast the value at 6 AM on the following day. Once the number and type of input and output variables are fixed, we implemented a forecasting model based on the neural network technique. We used a feedforward multilayer neural network and we trained it with the backpropagation algorithm. We tested networks with none, one and two hidden layers. The best model was one with one hidden layer, in contradiction with a previous study that found that minimum error was obtained with a net without hidden layer. Forecasts with the neural network are more accurate than those produced with a persistence model (the value six hours ahead is the same as the actual value). [ABSTRACT FROM AUTHOR]