• 산업공학
• /
• 제18권3호
• /
• pp.343-349
• /
• 2005

This study presents the empirical results of comparing several demand forecasting methods for Supply Chain Management(SCM). Adaptive exponential smoothing using change detection statistics (Jun) is compared with Trigg and Leach's adaptive methods and SAS time series forecasting systems using weekly SCM demand data. The results show that Jun's method is superior to others in terms of one-step-ahead forecast error and eight-step-ahead forecast error. Based on the results, we conclude that the forecasting performance of SCM solution can be improved by the proposed adaptive forecasting method.

• 한국멀티미디어학회논문지
• /
• 제20권2호
• /
• pp.363-370
• /
• 2017

Data assimilation is an initializing method for air quality forecasting such as PM10. It is very important to enhance the forecasting accuracy. Optimal interpolation is one of the data assimilation techniques. It is very effective and widely used in air quality forecasting fields. The technique, however, requires too much memory space and long execution time. It makes the PM10 air quality forecasting difficult in real time. We propose a fast optimal interpolation data assimilation method for PM10 air quality forecasting using a new kernel tridiagonal sparse matrix and CUDA massively parallel processing architecture. Experimental results show the proposed method is 5~56 times faster than conventional ones.

• 응용통계연구
• /
• 제32권1호
• /
• pp.69-81
• /
• 2019

본 논문에서는 딥러닝을 이용한 이변량 장기종속시계열(long-range dependent time series) 예측을 고려하였다. 시계열 데이터 예측에 적합한 LSTM(long short-term memory) 네트워크를 이용하여 이변량 장기종속시계열을 예측하고 이를 이변량 FARIMA(fractional ARIMA) 모형인 FIVARMA 모형과 VARFIMA 모형과의 예측 성능을 실증 자료 분석을 통해 비교하였다. 실증 자료로는 기능적 자기공명 영상(fMRI) 및 일일 실현 변동성(daily realized volatility) 자료를 이용하였으며 표본외 예측(out-of sample forecasting) 오차 비교를 통해 예측 성능을 측정하였다. 그 결과, FIVARMA 모형과 VARFIMA 모형의 예측값에는 미묘한 차이가 존재하며, LSTM 네트워크의 경우 초매개변수 선택으로 복잡해 보이지만 계산적으로 더 안정되면서 예측 성능도 모수적 장기종속시계열과 뒤지지 않은 좋은 예측 성능을 보였다.

• 한국수자원학회:학술대회논문집
• /
• 한국수자원학회 2023년도 학술발표회
• /
• pp.179-179
• /
• 2023

This study suggests a new approach of water level forecasting for extended lead times using original data preprocessing with variational mode decomposition (VMD). Here, two machine learning algorithms including light gradient boosting machine (LGBM) and random forest (RF) were considered to incorporate extended lead times (i.e., 5, 10, 15, 20, 25, 30, 40, and 50 days) forecasting of water levels. At first, the original data at two water level stations (i.e., SW173 and SW269 in Bangladesh) and their decomposed data from VMD were prepared on antecedent lag times to analyze in the datasets of different lead times. Mean absolute error (MAE), root mean squared error (RMSE), and mean squared error (MSE) were used to evaluate the performance of the machine learning models in water level forecasting. As results, it represents that the errors were minimized when the decomposed datasets were considered to predict water levels, rather than the use of original data standalone. It was also noted that LGBM produced lower MAE, RMSE, and MSE values than RF, indicating better performance. For instance, at the SW173 station, LGBM outperformed RF in both decomposed and original data with MAE values of 0.511 and 1.566, compared to RF's MAE values of 0.719 and 1.644, respectively, in a 30-day lead time. The models' performance decreased with increasing lead time, as per the study findings. In summary, preprocessing original data and utilizing machine learning models with decomposed techniques have shown promising results for water level forecasting in higher lead times. It is expected that the approach of this study can assist water management authorities in taking precautionary measures based on forecasted water levels, which is crucial for sustainable water resource utilization.

• 한국수자원학회:학술대회논문집
• /
• 한국수자원학회 2022년도 학술발표회
• /
• pp.141-141
• /
• 2022

Drought is a global phenomenon that affects almost all landscapes and causes major damages. Due to non-linear nature of contributing factors, drought occurrence and its severity is characterized as stochastic in nature. Early warning of impending drought can aid in the development of drought mitigation strategies and measures. Thus, drought forecasting is crucial in the planning and management of water resource systems. The primary objective of this study is to make improvement is existing drought forecasting techniques. Therefore, we proposed an improved version of Seasonal Autoregressive Integrated Moving Average (SARIMA) model (MD-SARIMA) for reliable drought forecasting with three years lead time. In this study, we selected four watersheds of Han River basin in South Korea to validate the performance of MD-SARIMA model. The meteorological data from 8 rain gauge stations were collected for the period 1973-2016 and converted into watershed scale using Thiessen's polygon method. The Standardized Precipitation Index (SPI) was employed to represent the meteorological drought at seasonal (3-month) time scale. The performance of MD-SARIMA model was compared with existing models such as Seasonal Naive Bayes (SNB) model, Exponential Smoothing (ES) model, Trigonometric seasonality, Box-Cox transformation, ARMA errors, Trend and Seasonal components (TBATS) model, and SARIMA model. The results showed that all the models were able to forecast drought, but the performance of MD-SARIMA was robust then other statistical models with Wilmott Index (WI) = 0.86, Mean Absolute Error (MAE) = 0.66, and Root mean square error (RMSE) = 0.80 for 36 months lead time forecast. The outcomes of this study indicated that the MD-SARIMA model can be utilized for drought forecasting.

• 한국시뮬레이션학회논문지
• /
• 제12권4호
• /
• pp.83-94
• /
• 2003

Managing a call center is a complex and diverse challenge. Call center becomes a very important contact point and a data ware house for successful CRM. Improving performance of call center is critical and valuable for providing better service. In this study we applied forecasting technique to estimate incoming calls and ProModel based simulation model to enhance performance of a mobile telecommunication company's call center. The simulation study shows reduction in managing cost and better customer's satisfaction.

• 산업공학
• /
• 제24권4호
• /
• pp.351-361
• /
• 2011

Forecasting of electricity demand have difficulty in adapting to abrupt weather changes along with a radical shift in major regional and global climates. This has lead to increasing attention to research on the immediate and accurate forecasting model. Technically, this implies that a model requires only a few input variables all of which are easily obtainable, and its predictive performance is comparable with other competing models. To meet the ends, this paper presents an energy demand forecasting model that uses the variable selection or extraction methods of data mining to select only relevant input variables, and employs support vector regression method for accurate prediction. Also, it proposes a novel performance measure for time-series prediction, shift index, followed by description on preprocessing procedure. A comparative evaluation of the proposed method with other representative data mining models such as an auto-regression model, an artificial neural network model, an ordinary support vector regression model was carried out for obtaining the forecast of monthly electricity demand from 2000 to 2008 based on data provided by Korea Energy Economics Institute. Among the models tested, the proposed method was shown promising results than others.

• Smart Structures and Systems
• /
• 제24권6호
• /
• pp.733-744
• /
• 2019

Wind speed forecasting is critical for a variety of engineering tasks, such as wind energy harvesting, scheduling of a wind power system, and dynamic control of structures (e.g., wind turbine, bridge, and building). Wind speed, which has characteristics of random, nonlinear and uncertainty, is difficult to forecast. Nowadays, machine learning approaches (generalized regression neural network (GRNN), back propagation neural network (BPNN), and extreme learning machine (ELM)) are widely used for wind speed forecasting. In this study, two schemes are proposed to improve the forecasting performance of machine learning approaches. One is that optimization algorithms, i.e., cross validation (CV), genetic algorithm (GA), and particle swarm optimization (PSO), are used to automatically find the optimal model parameters. The other is that the combination of different machine learning methods is proposed by finite mixture (FM) method. Specifically, CV-GRNN, GA-BPNN, PSO-ELM belong to optimization algorithm-assisted machine learning approaches, and FM is a hybrid machine learning approach consisting of GRNN, BPNN, and ELM. The effectiveness of these machine learning methods in wind speed forecasting are fully investigated by one-year field monitoring data, and their performance is comprehensively compared.

• Journal of Information Technology Applications and Management
• /
• 제30권3호
• /
• pp.1-13
• /
• 2023

Demand forecasting is a crucial task for an online retail where has to manage daily fresh foods effectively. Failing in forecasting results loss of profitability because of incompetent inventory management. This study investigated the optimal performance of different forecasting models for a very short shelf-life product. Demand data of 13 perishable items with aging of 210 days were used for analysis. Our comparison results of four methods: Trivial Identity, Seasonal Naïve, Feed-Forward and Autoregressive Recurrent Neural Networks (DeepAR) reveals that DeepAR outperforms with the lowest MAPE. This study also suggests the managerial implications by employing coefficient of variation (CV) as demand variation indicators. Three classes: Low, Medium and High variation are introduced for classify 13 products into groups. Our analysis found that DeepAR is suitable for medium and high variations, while the low group can use any methods. With this approach, the case can gain benefit of better fill-rate performance.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제18권1호
• /
• pp.1-14
• /
• 2024

Time-series forecasting is extensively used in the actual world. Recent research has shown that Transformers with a self-attention mechanism at their core exhibit better performance when dealing with such problems. However, most of the existing Transformer models used for time series prediction use the traditional encoder-decoder architecture, which is complex and leads to low model processing efficiency, thus limiting the ability to mine deep time dependencies by increasing model depth. Secondly, the secondary computational complexity of the self-attention mechanism also increases computational overhead and reduces processing efficiency. To address these issues, the paper designs an efficient multi-layer attention-based time-series forecasting model. This model has the following characteristics: (i) It abandons the traditional encoder-decoder based Transformer architecture and constructs a time series prediction model based on multi-layer attention mechanism, improving the model's ability to mine deep time dependencies. (ii) A cross attention module based on cross attention mechanism was designed to enhance information exchange between historical and predictive sequences. (iii) Applying a recently proposed sparse attention mechanism to our model reduces computational overhead and improves processing efficiency. Experiments on multiple datasets have shown that our model can significantly increase the performance of current advanced Transformer methods in time series forecasting, including LogTrans, Reformer, and Informer.