• The Korean Journal of Applied Statistics
• /
• v.28 no.5
• /
• pp.885-893
• /
• 2015

It is important to improve the forecasting accuracy of one-year-ahead seasonal factors in order to produce seasonally adjusted series of the following year. In this paper, seasonal factors of 8 monthly Korean economic time series are examined and forecast based on the functional principal component regression. One-year-ahead forecasts of seasonal factors from the functional principal component regression are compared with other forecasting methods based on mean absolute error (MAE) and mean absolute percentage error (MAPE). Forecasting seasonal factors via the functional principal component regression performs better than other comparable methods.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.18 no.3
• /
• pp.528-550
• /
• 2024

Warehousing demand prediction is an essential part of the supply chain, providing a fundamental basis for product manufacturing, replenishment, warehouse planning, etc. Existing forecasting methods cannot produce accurate forecasts since warehouse demand is affected by external factors such as holidays and seasons. Some aspects, such as consumer psychology and producer reputation, are challenging to quantify. The data can fluctuate widely or do not show obvious trend cycles. We introduce a new model for warehouse demand prediction called MAGRU, which stands for Multi-layer Attention with GRU. In the model, firstly, we perform the embedding operation on the input sequence to quantify the external influences; after that, we implement an encoder using GRU and the attention mechanism. The hidden state of GRU captures essential time series. In the decoder, we use attention again to select the key hidden states among all-time slices as the data to be fed into the GRU network. Experimental results show that this model has higher accuracy than RNN, LSTM, GRU, Prophet, XGboost, and DARNN. Using mean absolute error (MAE) and symmetric mean absolute percentage error(SMAPE) to evaluate the experimental results, MAGRU's MAE, RMSE, and SMAPE decreased by 7.65%, 10.03%, and 8.87% over GRU-LSTM, the current best model for solving this type of problem.

• Journal of The Korean Society of Agricultural Engineers
• /
• v.66 no.4
• /
• pp.51-57
• /
• 2024

This study was conducted on how to simulate runoff, which was done using existing physical models, using an LSTM (Long Short-Term Memory) model based on deep learning. Tancheon, the first tributary of the Han River, was selected as the target area for the model application. To apply the model, one water level observatory and four rainfall observatories were selected, and hourly data from 2020 to 2023 were collected to apply the model. River water level of the outlet of the Tancheon basin was simulated by inputting precipitation data from four rainfall observation stations in the basin and average preceding 72-hour precipitation data for each hour. As a result of water level simulation using 2021 to 2023 data for learning and testing with 2020 data, it was confirmed that reliable simulation results were produced through appropriate learning steps, reaching a certain mean absolute error in a short period time. Despite the short data period, it was found that the mean absolute percentage error was 0.5544~0.6226%, showing an accuracy of over 99.4%. As a result of comparing the simulated and observed values of the rapidly changing river water level during a specific heavy rain period, the coefficient of determination was found to be 0.9754 and 0.9884. It was determined that the performance of LSTM, which aims to simulate river water levels, could be improved by including preceding precipitation in the input data and using precipitation data from various rainfall observation stations within the basin.

• The Korean Journal of Applied Statistics
• /
• v.27 no.3
• /
• pp.419-430
• /
• 2014

An accurate prediction of electricity supply and demand is important for daily life, industrial activities, and national management. In this paper electricity sales is predicted by time series modelling. Real data analysis shows the transfer function model with cooling and heating days as an input time series and a pulse function as an intervention variable outperforms other time series models for the root mean square error and the mean absolute percentage error.

• Journal of Electrical Engineering and Technology
• /
• v.8 no.3
• /
• pp.559-564
• /
• 2013

In this paper a new hybrid forecast method composed of wavelet transform and neural network is proposed to forecast the wind speed more accurately. In the field of wind energy research, accurate forecast of wind speed is a challenging task. This will influence the power system scheduling and the dynamic control of wind turbine. The wind data used here is measured at 15 minute time intervals. The performance is evaluated based on the metrics, namely, mean square error, mean absolute error, sum squared error of the proposed model and compared with the back propagation model. Simulation studies are carried out and it is reported that the proposed model outperforms the compared model based on the metrics used and conclusions were drawn appropriately.

• Journal of The Korean Society of Agricultural Engineers
• /
• v.51 no.3
• /
• pp.53-62
• /
• 2009

This study was carried out to select optimal probability distribution based on design accumulated monthly mean inflow from the viewpoint of drought by Gamma (GAM), Generalized extreme value (GEV), Generalized logistic (GLO), Generalized normal (GNO), Generalized pareto (GPA), Gumbel (GUM), Normal (NOR), Pearson type 3 (PT3), Wakeby (WAK) and Kappa (KAP) distributions for the observed accumulative monthly mean inflow of Chungjudam. L-moment ratio was calculated using observed accumulative monthly mean inflow. Parameters of 10 probability distributions were estimated by the method of L-moments with the observed accumulated monthly mean inflow. Design accumulated monthly mean inflows obtained by the method of L-moments using different methods for plotting positions formulas in the 10 probability distributions were compared by relative mean error (RME) and relative absolute error (RAE) respectively. It has shown that the design accumulative monthly mean inflow derived by the method of L-moments using Weibull plotting position formula in WAK and KAP distributions were much closer to those of the observed accumulative monthly mean inflow in comparison with those obtained by the method of L-moment with the different formulas for plotting positions in other distributions from the viewpoint of RME and RAE.

• Communications for Statistical Applications and Methods
• /
• v.10 no.3
• /
• pp.637-645
• /
• 2003

Small area estimation had been studied using data-based methods such as Direct, Indirect, Synthetic methods. However recently, model-based such as based on regression or time series estimation methods are applied to the study. In this paper we investigate a model-data based small area estimation which takes into account the spatial relation among the areas. The Economic Active Population Survey in 2001 are used for analysis and the results from the model based and model-data based estimation are compared with using MSE(Mean squared error), MAE(Mean absolute error) and MB(Mean bias).

• Journal of the Korean Operations Research and Management Science Society
• /
• v.23 no.1
• /
• pp.67-95
• /
• 1998

This research addresses flow time prediction in the dynamic unbalanced job shop scheduling environment. The specific purpose of the research is to develop the job flow time prediction model in the dynamic unbalance djob shop. Such factors as job characteristics, job shop status, characteristics of the shop workload, shop dispatching rules, shop structure, etc, are considered in the prediction model. The regression prediction approach is analyzed within a dynamic, make-to-order job shop simulation model. Mean Absolute Lateness (MAL) and Mean Relative Error (MRE) are used to compare and evaluate alternative regression models devloped in this research.

• Transactions of Materials Processing
• /
• v.29 no.5
• /
• pp.282-289
• /
• 2020

Flexible roll forming is an advanced sheet-metal-forming process that allows the production of parts with various cross-sections. During the flexible process, material is subjected to three-dimensional deformation such as transverse bending, inhomogeneous elongations, or contraction. Because of the effects of process variables on the quality of the roll-formed products, the approaches used to investigate the roll-forming process have been largely dependent on experience and trial- and-error methods. Web-warping is one of the major shape defects encountered in flexible roll forming. In this study, an SVR model was developed to predict the web-warping during the flexible roll forming process. In the development of the SVR model, three process parameters, namely the forming-roll speed condition, leveling-roll height, and bend angle were considered as the model inputs, and the web-warping height was used as the response variable for three blank shapes; rectangular, concave, and convex shape. MATLAB software was used to train the SVR model and optimize three hyperparameters (λ, ε, and γ). To evaluate the SVR model performance, the statistical analysis was carried out based on the three indicators: the root-mean-square error, mean absolute error, and relative root-mean-square error.

• Korea Trade Review
• /
• v.47 no.3
• /
• pp.211-232
• /
• 2022

Due to the impact of the public health event COVID-19 epidemic, the Chinese futures market showed "Black Swan". This has brought the unpredictable into the economic environment with many commodities falling by the daily limit, while gold performed well and closed in the sunshine(Yan-Li and Rui Qian-Wang, 2020). Volatility is integral part of financial market. As an emerging market and a special precious metal, it is important to forecast return of gold futures price. This study selected data of the SHFE gold futures returns and conducted an empirical analysis based on the generalised autoregressive conditional heteroskedasticity (GARCH)-type model. Comparing the statistics of AIC, SC and H-QC, ARMA (12,9) model was selected as the best model. But serial correlation in the squared returns suggests conditional heteroskedasticity. Next part we established the autoregressive moving average ARMA-GARCH-type model to analysis whether Volatility Clustering and the leverage effect exist in the Chinese gold futures market. we consider three different distributions of innovation to explain fat-tailed features of financial returns. Additionally, the error degree and prediction results of different models were evaluated in terms of mean squared error (MSE), mean absolute error (MAE), Theil inequality coefficient(TIC) and root mean-squared error (RMSE). The results show that the ARMA(12,9)-TGARCH(2,2) model under Student's t-distribution outperforms other models when predicting the Chinese gold futures return series.