• Journal of Korea Port Economic Association
• /
• v.26 no.1
• /
• pp.222-233
• /
• 2010

This paper aims at predicting the BDI from Jan. to Dec. 2010 using such econometric techniues of the univariate time series as stochastic ARIMA-type models and Hodrick-Prescott filtering technique. The multivariate cause-effect econometric model is not employed for not assuring a higher degree of forecasting accuracy than the univariate variable model. Such a cause-effect econometric model also fails in adjusting itself for the post-sample. This article introduces the two ARIMA models and five Intervention-ARIMA models. The monthly data cover the period January 2000 through December 2009. The out-of-sample forecasting performance is compared between the ARIMA-type models and the random walk model. Forecasting performance is measured by three summary statistics: root mean squared error (RMSE), mean absolute error (MAE) and mean error (ME). The RMSE and MAE indicate that the ARIMA-type models outperform the random walk model And the mean errors for all models are small in magnitude relative to the MAE's, indicating that all models don't have a tendency of overpredicting or underpredicting systematically in forecasting. The pessimistic ex-ante forecasts are expected to be 2,820 at the end of 2010 compared with the optimistic forecasts of 4,230.

• Journal of Korea Technology Innovation Society
• /
• v.14 no.spc
• /
• pp.1232-1255
• /
• 2011

This paper proposes a model for demand forecasting that will require less effort in the process of utilizing the new product diffusion model while also allowing for more objective and timely application. Drawing upon the theoretical foundation provided by the hype cycle model and the consumer adoption model, this proposed model makes it possible to estimate the maximum market potential based solely on bibliometrics and the scale of the early market, thereby presenting a method for supplying the major parameters required for the Bass model. Upon analyzing the forecasting ability of this model by applying it to the case of the hybrid car market, the model was confirmed to be capable of successfully forecasting results similar in scale to the market potential deduced through various other objective sources of information, thus underscoring the potentials of utilizing this model. Moreover, even the hype cycle or the life cycle can be estimated through direct linkage with bibliometrics and the Bass model. In cases where the hype cycles of other models have been observed, the forecasting ability of this model was demonstrated through simple case studies. Since this proposed model yields a maximum market potential that can also be applied directly to other growth curve models, the model presented in the following paper provides new directions in the endeavor to forecast technology diffusion and identify promising technologies through bibliometrics.

• The Korean Journal of Applied Statistics
• /
• v.32 no.1
• /
• pp.69-81
• /
• 2019

We consider bivariate long range dependent (LRD) time series forecasting using a deep learning method. A long short-term memory (LSTM) network well-suited to time series data is applied to forecast bivariate time series; in addition, we compare the forecasting performance with bivariate fractional autoregressive integrated moving average (FARIMA) models. Out-of-sample forecasting errors are compared with various performance measures for functional MRI (fMRI) data and daily realized volatility data. The results show a subtle difference in the predicted values of the FIVARMA model and VARFIMA model. LSTM is computationally demanding due to hyper-parameter selection, but is more stable and the forecasting performance is competitively good to that of parametric long range dependent time series models.

• The Korean Journal of Applied Statistics
• /
• v.36 no.2
• /
• pp.129-139
• /
• 2023

Forecasting volatility is essential to avoiding the risk caused by the uncertainties of an financial asset. Complicated financial volatility features such as ambiguity between non-stationarity and stationarity, asymmetry, long-memory, sudden fairly large values like outliers bring great challenges to volatility forecasts. In order to address such complicated features implicity, we consider machine leaning models such as LSTM (1997) and GRU (2014), which are known to be suitable for existing time series forecasting. However, there are the problems of vanishing gradients, of enormous amount of computation, and of a huge memory. To solve these problems, a causal temporal convolutional network (TCN) model, an advanced form of 1D CNN, is also applied. It is confirmed that the overall forecasting power of TCN model is higher than that of the RNN models in forecasting VIX, VXD, and VXN, the daily volatility indices of S&P 500, DJIA, Nasdaq, respectively.

• Journal of the Korean Operations Research and Management Science Society
• /
• v.26 no.4
• /
• pp.55-70
• /
• 2001

Artificial neural network(ANN) models have been widely used for the classification problems in business such as bankruptcy prediction, credit evaluation, etc. Although the application of ANN to classification of consumers＇ choice behavior is a promising research area, there have been only a few researches. In general, most of the researches have reported that the classification performance of the ANN models were better than conventional statistical model Because the survey data on consumer behavior may include much noise and missing data, ANN model will be more robust than conventional statistical models welch need various assumptions. The purpose of this paper is to study the potential of the ANN model for forecasting consumers＇ choice behavior based on survey data. The data was collected by questionnaires to the shoppers of department stores and discount stores. Then the correct classification rates of the ANN models for the training and test sample with that of multiple discriminant analysis(MDA) and logistic regression(Logit) model. The performance of the ANN models were betted than the performance of the MDA and Logit model with respect to correct classification rate. By using input variables identified as significant in the stepwise MDA, the performance of the ANN models were improved.

• The Bulletin of The Korean Astronomical Society
• /
• v.42 no.2
• /
• pp.60.2-61
• /
• 2017

A Convolutional Neural Network(CNN) is one of the well-known deep-learning methods in image processing and computer vision area. In this study, we apply CNN to two kinds of flare forecasting models: flare classification and occurrence. For this, we consider several pre-trained models (e.g., AlexNet, GoogLeNet, and ResNet) and customize them by changing several options such as the number of layers, activation function, and optimizer. Our inputs are the same number of SOHO)/MDI images for each flare class (None, C, M and X) at 00:00 UT from Jan 1996 to Dec 2010 (total 1600 images). Outputs are the results of daily flare forecasting for flare class and occurrence. We build, train, and test the models on TensorFlow, which is well-known machine learning software library developed by Google. Our major results from this study are as follows. First, most of the models have accuracies more than 0.7. Second, ResNet developed by Microsoft has the best accuracies : 0.77 for flare classification and 0.83 for flare occurrence. Third, the accuracies of these models vary greatly with changing parameters. We discuss several possibilities to improve the models.

• Korea Trade Review
• /
• v.46 no.6
• /
• pp.191-209
• /
• 2021

In this paper, various time series analysis models and machine learning models are presented for long-term prediction of export growth rate, and the prediction performance is compared and reviewed by RMSE and MAE. Export growth rate is one of the major economic indicators to evaluate the economic status. And It is also used to predict economic forecast. The export growth rate may have a negative (-) value as well as a positive (+) value. Therefore, Instead of using the ReLU function, which is often used for time series prediction of deep learning models, the PReLU function, which can have a negative (-) value as an output value, was used as the activation function of deep learning models. The time series prediction performance of each model for three types of data was compared and reviewed. The forecast data of long-term prediction of export growth rate was deduced by three forecast methods such as a fixed forecast method, a recursive forecast method and a rolling forecast method. As a result of the forecast, the traditional time series analysis model, ARDL, showed excellent performance, but as the time period of learning data increases, the performance of machine learning models including LSTM was relatively improved.

• The Transactions of The Korean Institute of Electrical Engineers
• /
• v.63 no.11
• /
• pp.1497-1502
• /
• 2014

Electric load forecasting is essential for stable electric power supply, efficient operation and management of power systems, and safe operation of power generation systems. The results are utilized in generator preventive maintenance planning and the systemization of power reserve management. Development and improvement of electric load forecasting model is necessary for power system maintenance and operation. This paper proposes daily maximum electric load forecasting methods for the next 4 weeks with a seasonal autoregressive integrated moving average model and an exponential smoothing model. According to the results of forecasting of daily maximum electric load forecasting for the next 4 weeks of March, April, November 2010~2012 using the constructed forecasting models, the seasonal autoregressive integrated moving average model showed an average error rate of 6,66%, 5.26%, 3.61% respectively and the exponential smoothing model showed an average error rate of 3.82%, 4.07%, 3.59% respectively.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.15 no.4
• /
• pp.437-442
• /
• 2005

The forecasting of flood runoff in the river is essential for flood control. The purpose of this study is to test a development of system for flood runoff forecasting using neural network model. For the flood events the tested rainfall and runoff data were the input to the input layer and the flood runoff data were used in the output layer To choose the forecasting model which would make up of runoff forecasting system properly, real-time runoff in the river when flood periods were forecasted by using the neural network model and the state-space model. A comparison of the results obtained by the two forecasting models indicated the superiority and reliability of the neural network model over the state-space model. The neural network model was modified to work in the Web and developed to be the basic model of the forecasting system for the flood runoff.

• Journal of the Korean Institute of Illuminating and Electrical Installation Engineers
• /
• v.28 no.1
• /
• pp.50-56
• /
• 2014

Accurate midterm load forecasting is essential to preventive maintenance programs and reliable demand supply programs. This paper describes a midterm load forecasting method using autoregressive integrated moving average (ARIMA) model which has been widely used in time series forecasting due to its accuracy and predictability. The various ARIMA models are examined in order to find the optimal model having minimum error of the midterm load forecasting. The proposed method is applied to forecast 104-week load pattern using the historical data in Korea. The effectiveness of the proposed method is evaluated by forecasting 104-week load from 2011 to 2012 by using historical data from 2002 to 2010.