• Industrial Engineering and Management Systems
• /
• 제14권2호
• /
• pp.196-209
• /
• 2015

Retail sales forecast is a special area of forecasting. Its unique characteristics call for unique data models and treatment, and unique forecasting processes. In this paper, we will address lessons learned and challenges encountered in retail sales forecast from a practical and technical perspective. In particular, starting with the data models of retail sales data, we proceed to address issues existing in estimating and processing each component in the data model. We will discuss how to estimate the multi-seasonal cycles in retail sales data, and the limitations of the existing methodologies. In addition, we will talk about the distinction between business events and forecast events, the methodologies used in event detection and event effect estimation, and the difficulties in compound event detection and effect estimation. For each of the issues and challenges, we will present our solution strategy. Some of the solution strategies can be generalized and could be helpful in solving similar forecast problems in different areas.

• 한국경영과학회:학술대회논문집
• /
• 대한산업공학회/한국경영과학회 1996년도 춘계공동학술대회논문집; 공군사관학교, 청주; 26-27 Apr. 1996
• /
• pp.741-744
• /
• 1996

Artifical Neural Network(ANN) models were used for forecasting interest rate as a new methodology, which has proven itself successful in financial domain. This research intended to construct ANN models which can maximize the performance of prediction, regarding Corporate Bond Yield (CBY) as interest rate. Synergistic Market Analysis (SMA) was applied to the construction of models [Freedman et al.]. In this aspect, while the models which consist of only time series data for corporate bond yield were devloped, the other models generated through conjunction and reorganization of fundamental variables and market variables were developed. Every model was constructed to predict 1,6, and 12 months after and we obtained 9 ANN models for interest rate forecasting. Multi-layer perceptron networks using backpropagation algorithm showed good performance in the prediction for 1 and 6 months after.

• Industrial Engineering and Management Systems
• /
• 제7권1호
• /
• pp.9-22
• /
• 2008

As an extension of a previous work by the authors (Song and Esogbue, 2006), a new algorithm for automated modeling of nonstationary seasonal time series is presented in this paper. Issues relative to the methodology for building automatically seasonal time series models and periodic time series models are addressed. This is achieved by inspecting the trend, estimating the seasonality, determining the orders of the model, and estimating the parameters. As in our previous work, the major instruments used in the model identification process are correlograms of the modeling errors while the least square method is used for parameter estimation. We provide numerical illustrations of the performance of the new algorithms with respect to building both seasonal time series and periodic time series models. Additionally, we consider forecasting and exercise the models on some sample time series problems found in the literature as well as real life problems drawn from the retail industry. In each instance, the models are built automatically avoiding the necessity of any human intervention.

• 생태와환경
• /
• 제46권1호
• /
• pp.1-9
• /
• 2013

Algal blooms in potable water supplies are becoming an increasingly prevalent and serious water quality problem around the world. In addition to precipitating taste and odor problems, blooms damage the environment, and some classes like cyanobacteria (blue-green algae) release toxins that can threaten human health, even causing death. There is a recognized need in the water industry for models that can accurately forecast in real-time algal bloom events for planning and mitigation purposes. In this study, using data for an interconnected system of rivers and reservoirs operated by a New Jersey water utility, various ANN models, including both discrete prediction and classification models, were developed and tested for forecasting counts of three different algal classes for one-week and two-weeks ahead periods. Predictor model inputs included physical, meteorological, chemical, and biological variables, and two different temporal schemes for processing inputs relative to the prediction event were used. Despite relatively limited historical data, the discrete prediction ANN models generally performed well during validation, achieving relatively high correlation coefficients, and often predicting the formation and dissipation of high algae count periods. The ANN classification models also performed well, with average classification percentages averaging 94 percent accuracy. Despite relatively limited data events, this study demonstrates that with adequate data collection, both in terms of the number of historical events and availability of important predictor variables, ANNs can provide accurate real-time forecasts of algal population counts, as well as foster increased understanding of important cause and effect relationships, which can be used to both improve monitoring programs and forecasting efforts.

• 한국수자원학회논문집
• /
• 제34권6호
• /
• pp.701-711
• /
• 2001

수질 인자들은 다양하고 관계가 복잡하여 수질 변화를 예측하는데 많은 어려움이 있다. 따라서 입력과 출력이 비교적 용이하고 비선형 예측에 적합한 신경망 모형을 이용하여 금강유역 공주지점의 DO, BOD, TN에 대한 월수질 예측을 수행하고 ARIMA 모형과 비교하여 적용 가능성을 검토하였다. 사용된 신경망 모형은 학습을 위해 BP(Back Propagation) 알고리즘을 적용하였으며 학습을 향상시키기 위한 모멘트-적응학습율(Moment-Adaptive learming rate) 방법을 이용한 MANN 모형, 레번버그-마쿼트(Levenberg-Marquardt) 방법을 이 용한 LMNN 모형, 그리고 정성적인 판단인자를 첨가하여 정량적인 월 수질 자료와 분별, 학습하 도록 은닉층을 분리한 MNN 모형으로 구분하였다. 대체로 신경망 모형의 예측치가 실측치에 근사한 결과를 보였으며, 은닉층을 분리한 MNN 모형이 가장 우수한 결과를 보였다.

• Asia pacific journal of information systems
• /
• 제7권1호
• /
• pp.67-83
• /
• 1997

The utility of learning techniques in investment analysis has been demonstrated in many areas, ranging from forecasting individual stocks to entire market indexes. To date, however, the application of artificial intelligence to financial forecasting has focused largely on short predictive horizons. Usually the forecast window is a single period ahead; if the input data involve daily observations, the forecast is for one day ahead; if monthly observations, then a month ahead; and so on. Thus far little work has been conducted on the efficacy of long-term prediction involving multiperiod forecasting. This paper examines the impact of alternative procedures for extended prediction using knowledge discovery techniques. One dimension in the study involves temporal granularity: a single jump from the present period to the end of the forecast window versus a web of short-term forecasts involving a sequence of single-period predictions. Another parameter relates to the numerosity of input variables: a technical approach involving only lagged observations of the target variable versus a fundamental approach involving multiple variables. The dual possibilities along each of the granularity and numerosity dimensions entail a total of 4 models. These models are first evaluated using neural networks, then compared against a multi-input jump model using case based reasoning. The computational models are examined in the context of forecasting the S&P 500 index.

• 산업공학
• /
• 제24권4호
• /
• pp.373-378
• /
• 2011

High level of accuracy in forecasting heat demand of each district is required for operating and managing the district heating efficiently. Heat demand has a close connection with the demands of the previous days and the temperature, general demand forecasting methods may be used forecast. However, there are some exceptional situations to apply general methods such as the exceptional low demand in weekends or vacation period. We introduce a new method to forecast the heat demand to overcome these situations, using the linearities between the demand and some other factors. Our method uses the temperature and the past 7 days' demands as the factors which determine the future demand. The model consists of daily and hourly models which are multiple linear regression models. Appling these two models to historical data, we confirmed that our method can forecast the heat demand correctly with reasonable errors.

• 상하수도학회지
• /
• 제19권6호
• /
• pp.738-748
• /
• 2005

Considering pollution source is transferred by discharge, it is very important to analyze the correlation between discharge and water quality. And temperature also influent to the water quality. In this paper, it is used water quality data that was measured DO (Dissolved Oxygen), TOC (Total Organic Carbon), TN (Total Nitrogen), TP (Total Phosphorus) at Dalchun real time monitoring stations in Namhan river. These characteristics were analyzed with the water quality of rainy and nonrainy periods. Input data of the water quality forecasting models that they were constructed by neural network and neuro-fuzzy was chosen as the reasonable data, and water quality forecasting models were applied. LMNN (Levenberg-Marquardt Neural Network), MDNN (MoDular Neural Network), and ANFIS (Adaptive Neuro-Fuzzy Inference System) models have achieved the highest overall accuracy of TOC data. LMNN and MDNN model which are applied for DO, TN, TP forecasting shows better results than ANFIS. MDNN model shows the lowest estimation error when using daily time, which is qualitative data trained with quantitative data. If some data has periodical properties, it seems effective using qualitative data to forecast.

• 대한전기학회논문지:전력기술부문A
• /
• 제49권3호
• /
• pp.107-117
• /
• 2000

This paper proposes a systematic method to develop short-term electrical load forecasting systems using neuro-fuzzy models. The primary goal of the proposed method is to improve the performance of the prediction model in terms of accuracy and reliability. For this, the proposed method explores the advantages of the structure learning of the neuro-fuzzy model. The proposed load forecasting system first builds an initial structure off-line for each hour of four day types and then stores the resultant initial structures in the initial structure bank. Whenever a prediction needs to be made, the proposed system initializes the neuro-fuzzy model with the appropriate initial structure stored and trains the initialized model. In order to demonstrate the viability of the proposed method, we develop an one hour ahead load forecasting system by using the real load data collected during 1993 and 1994 at KEPCO. Simulation results reveal that the prediction system developed in this paper can achieve a remarkable improvement on both accuracy and reliability compared with the prediction systems based on multilayer perceptrons, radial basis function networks, and neuro-fuzzy models without the structure learning.

• Nutritional Sciences
• /
• 제9권3호
• /
• pp.201-211
• /
• 2006

The purpose of this study was to provide dietitians with the guidance in forecasting meal counts for a university/college foodservice facility. The forecasting methods to be analyzed were the following: naive model 1, 2, and 3; moving average, double moving average, simple exponential smoothing, double exponential smoothing, Holt's, and Winters' methods, and simple linear regression. The accuracy of the forecasting methods was measured using mean squared error and Theil's U-statistic. This study showed how to project meal counts using 10 forecasting methods for dietitians. The results of this study showed that WES was the most accurate forecasting method, followed by $na\ddot{i}ve$ 2 and naive 3 models. However, naive model 2 and 3 were recommended for using by dietitians in university/college dining facilities because of the accuracy and ease of use. In addition, the 2000 spring semester data were better than the 2000 fall semester data to forecast 2001spring semester data.