A hybrid SARIMA wavelet transform method for sales forecasting

Choi, Tsan‐Ming; Yu, Yong; Au, Kin Fan

doi:10.1016/j.dss.2010.12.002

Cited by 133 publications

(50 citation statements)

References 17 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The proposed approach consists of two phases, viz. decomposition and reconstruction [8,15]. In the first phase, the series is decomposed into high (detailed) and low (approximate) pass filters, which respectively pick up the higher and lower frequency components of the series.…”

Section: The Proposed Dwt Based Arima-ann Hybrid Methodsmentioning

confidence: 99%

“…Milidiu et al [7] used haar wavelet together with a clustering algorithm to partition input data into different regions. Choi et al [8] used a hybrid SARIMA and wavelet transform to forecast the sales time series. Conejo et al [9] used wavelet transform to decompose electricity price time series and then applied ARIMA to forecast future prices.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Time Series Forecasting Using Hybrid ARIMA and ANN Models Based on DWT Decomposition

Khandelwal

Adhikari

Verma

2015

Procedia Computer Science

237

View full text Add to dashboard Cite

Recently Discrete Wavelet Transform (DWT) has led to a tremendous surge in many domains of science and engineering. In this study, we present the advantage of DWT to improve time series forecasting precision. This article suggests a novel technique of forecasting by segregating a time series dataset into linear and nonlinear components through DWT. At first, DWT is used to decompose the in-sample training dataset of the time series into linear (detailed) and non-linear (approximate) parts. Then, the Autoregressive Integrated Moving Average (ARIMA) and Artificial Neural Network (ANN) models are used to separately recognize and predict the reconstructed detailed and approximate components, respectively. In this manner, the proposed approach tactically utilizes the unique strengths of DWT, ARIMA, and ANN to improve the forecasting accuracy. Our hybrid method is tested on four real-world time series and its forecasting results are compared with those of ARIMA, ANN, and Zhang's hybrid models. Results clearly show that the proposed method achieves best forecasting accuracies for each series.

show abstract

Section: The Proposed Dwt Based Arima-ann Hybrid Methodsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Time Series Forecasting Using Hybrid ARIMA and ANN Models Based on DWT Decomposition

Khandelwal

Adhikari

Verma

2015

Procedia Computer Science

237

View full text Add to dashboard Cite

show abstract

“…Looking at the original data, many time series display the patterns of significant autocorrelation and annual cycles. It is important to note that the detail components of signal decomposition in WT can be associated with factors such as seasonal cycles, and other influencing variables that may be external to the time series (Choi et al, 2011). Therefore, the correlograms of the detail components were used to check whether any cyclical patterns were still present post-decomposition.…”

Section: Seasonality Factormentioning

confidence: 99%

Using discrete wavelet transforms to analyze trends in streamflow and precipitation in Quebec and Ontario (1954–2008)

Nalley¹,

Adamowski²,

Khalil³

2012

Journal of Hydrology

252

149

View full text Add to dashboard Cite

“…Choi et al [11] and Kutluk et al [12] both proposed the classic SARIMA method for load forecasting while James Taylor extended double seasonal ARMA model which includes intraday and intraweek seasonal cycles to include intrayear seasonal cycle, which is also apparent if one disposes of a multi-year training dataset. Weather features were also used to construct a classic ARMA/SARIMA model, which can be found in Jennifer et al's work.…”

Section: Introductionmentioning

confidence: 99%

Short-Term Forecasting of Temperature Driven Electricity Load Using Time Series and Neural Network Model

Liu¹,

Babushkin²,

Afshari³

2014

JOCET

View full text Add to dashboard Cite

Abstract-In this paper, two models were proposed for week-ahead forecasting of temperature driven electricity load, which are a time series model and an Artificial Neural Network (ANN) model. Over the week-long ("future") forecasting horizon, predicted temperature from ANN was used as it is shown that ANN produced more accurate temperature prediction. For the time series model, Seasonal Autoregressive Integrated Moving Average with eXogenous variables (SARIMAX) scheme was proposed. A method called "pre-whitening" was used to determine the lagged effect of temperature on electricity load. Comparison between ANN model and SARIMAX model was conducted to see which one gave a better forecasting performance. The forecast performance was characterized by two statistical estimates, the Root Mean Square Error (RMSE), Mean Absolute Percentage Error (MAPE). The results showed that while the ANN model behaved better in the estimation stage, its performance got worse than SARIMAX model in the forecasting stage.Index Terms-Artificial neural networks (ANN), load forecasting, SARIMAX, short-term, temperature forecasting, time series. I. INTRODUCTIONAfter the deregulation of electricity markets, electricity was commoditized. As a result, the generation of electricity more flexible and demand oriented. However, there are also risks associated the deregulation of electricity markets such as electricity oversupply and shortage due to inaccurate forecasting, which could result in significant financial loss. That is why accurate electricity forecasting plays a very important role and could also improve power generation planning. In this study two kinds of models, SARIMAX and ANN, were proposed for short-term forecasting of temperature driven electricity load forecasting.Different approaches have been proposed for the short-term forecasting of electricity load. Generally speaking, these approaches can be grouped into three categories: regression-based, time series, artificial intelligence and computational intelligence. The latter can divided into several sub-groups, such as neural networks, support vector machines, hybrid and other approaches. In the following section, mainly neural networks and time series approaches will be studied from the literature.Ghanbari et al. et al. [9] and Martí nez-Álvarez [10] all proposed an approach based on selection of similar days according to which the load curves are forecasted by using the information of the days being similar to that of the forecast day.Choi et al. [11] and Kutluk et al. [12] both proposed the classic SARIMA method for load forecasting while James Taylor extended double seasonal ARMA model which includes intraday and intraweek seasonal cycles to include intrayear seasonal cycle, which is also apparent if one disposes of a multi-year training dataset. Weather features were also used to construct a classic ARMA/SARIMA model, which can be found in Jennifer et al. 's work. [5] G. Peter [13] proposed a hybrid methodology that combines both ARIMA and ANN models to take advantage ...

show abstract

A hybrid SARIMA wavelet transform method for sales forecasting

Cited by 133 publications

References 17 publications

Time Series Forecasting Using Hybrid ARIMA and ANN Models Based on DWT Decomposition

Time Series Forecasting Using Hybrid ARIMA and ANN Models Based on DWT Decomposition

Using discrete wavelet transforms to analyze trends in streamflow and precipitation in Quebec and Ontario (1954–2008)

Short-Term Forecasting of Temperature Driven Electricity Load Using Time Series and Neural Network Model

Contact Info

Product

Resources

About