Neural network forecasting for seasonal and trend time series

Zhang, G.Peter; Qi, Meiling

doi:10.1016/j.ejor.2003.08.037

Cited by 764 publications

(440 citation statements)

References 25 publications

Supporting

Mentioning

385

Contrasting

Unclassified

Order By: Relevance

“…We compared our results with that of results obtained from the hybrid ARIMA-NN model of Zhang and Qi [2]. Here, they use a first order polynomial fit to de-trend, and an ARIMA variant to de-seasonalise the data.…”

Section: Methodmentioning

confidence: 76%

“…In contrast, pre-processing techniques can be used to stabilise the mean and variance of a series [6], but the choice of technique is controversial [2]. We describe a new procedure for stabilising a time series for which the choice of pre-processing technique is optimised for a TDNN.…”

mentioning

confidence: 99%

“…Furthermore, as for other types of neural network, a TDNN cannot predict trend because the output of the network is typically computed using a threshold function, with the output of the function bounded to values, say, between 0 and 1 [7]. Autoregressive models, such as the seasonal autoregressive integrated moving average model (SARIMA), have been used to generate deseasonalised and de-trended time series, with the residuals used to train a TDNN, obtaining good results [2,3,4]. In this letter we present a pre-processing method that can be used to stabilise a time series appropriately for the application of a TDNN.…”

mentioning

confidence: 99%

See 2 more Smart Citations

Pre-processing inputs for optimally-configured time-delay neural networks

2005

View full text Add to dashboard Cite

A procedure for pre-processing non-stationary time series is proposed for modelling with a time-delay neural network (TDNN). The procedure stabilises the mean of the series and uses a Fast Fourier Transform (FFT) to determine the TDNN input size. Results of applying this procedure on five well-known data sets are compared with existing hybrid neural network techniques, demonstrating improved prediction performance.

show abstract

Section: Methodmentioning

confidence: 76%

mentioning

confidence: 99%

mentioning

confidence: 99%

See 1 more Smart Citation

Pre-processing inputs for optimally-configured time-delay neural networks

2005

View full text Add to dashboard Cite

show abstract

“…As the second algorithm, the Kriging method has previously been used in environmental estimation problems defined within a continuous feature space (such as temperature inside a homogeneous container) [48]. Finally, as the third algorithm, we assume that the inherently nonlinear relationship between the product temperature inside a pallet and the air temperature can be modelled by an artificial neural network (ANN) as shown in figure 10 [50]. Input to the network consists of time-temperature data provided by the sensor(s) placed outside the pallet, whereas output is the estimated time-temperature data for products placed inside the pallet.…”

Section: How To Estimate Product Temperatures With External Pallet Sementioning

confidence: 99%

Improvement in fresh fruit and vegetable logistics quality: berry logistics field studies

Nunes

Nicometo

Émond

et al. 2014

Phil. Trans. R. Soc. A.

View full text Add to dashboard Cite

Shelf life of fresh fruits and vegetables is greatly influenced by environmental conditions. Increasing temperature usually results in accelerated loss of quality and shelf-life reduction, which is not physically visible until too late in the supply chain to adjust logistics to match shelf life. A blackberry study showed that temperatures inside pallets varied significantly and 57% of the berries arriving at the packinghouse did not have enough remaining shelf life for the longest supply routes. Yet, the advanced shelf-life loss was not physically visible. Some of those pallets would be sent on longer supply routes than necessary, creating avoidable waste. Other studies showed that variable pre-cooling at the centre of pallets resulted in physically invisible uneven shelf life. We have shown that using simple temperature measurements much waste can be avoided using ‘first expiring first out’. Results from our studies showed that shelf-life prediction should not be based on a single quality factor as, depending on the temperature history, the quality attribute that limits shelf life may vary. Finally, methods to use air temperature to predict product temperature for highest shelf-life prediction accuracy in the absence of individual sensors for each monitored product have been developed. Our results show a significant reduction of up to 98% in the root-mean-square-error difference between the product temperature and air temperature when advanced estimation methods are used.

show abstract

“…There is abundant literature on nonlinear models for time series forecasting [3,4,5,6,7,8]. Among the existing methods are neural networks [9,10,11,12,13,14,15], radial basis function networks [11,16,17,18], support vector machines [19,20,21,22], self organizing maps [23,24] and other variants of these models [11,25,26,27,28]. However, building these models takes considerable computational time compared to linear models.…”

Section: Introductionmentioning

confidence: 99%

Approximate k-NN delta test minimization method using genetic algorithms: Application to time series

2010

View full text Add to dashboard Cite

In many real world problems, the existence of irrelevant input variables (features) hinders the predictive quality of the models used to estimate the output variables. In particular, time series prediction often involves building large regressors of artificial variables that can contain irrelevant or misleading information. Many techniques have arisen to confront the problem of accurate variable selection, including both local and global search strategies. This paper presents a method based on genetic algorithms that intends to find a global optimum set of input variables that minimize the Delta Test criterion. The execution speed has been enhanced by substituting the exact nearest neighbor computation by its approximate version. The problems of scaling and projection of variables have been addressed. The developed method works in conjunction with MATLAB's Genetic Algorithm and Direct Search Toolbox. The goodness of the proposed methodology has been evaluated on several popular time series examples, and also generalized to other non-time-series datasets.

show abstract

Neural network forecasting for seasonal and trend time series

Cited by 764 publications

References 25 publications

Pre-processing inputs for optimally-configured time-delay neural networks

Pre-processing inputs for optimally-configured time-delay neural networks

Improvement in fresh fruit and vegetable logistics quality: berry logistics field studies

Approximate k-NN delta test minimization method using genetic algorithms: Application to time series

Contact Info

Product

Resources

About