• The Korean Journal of Applied Statistics
• /
• v.32 no.6
• /
• pp.851-865
• /
• 2019

Discovering various features from one time series is complicated. In this paper, we introduce a multi aggregation prediction algorithm (MAPA) that uses the concepts of temporal aggregation and combining forecasts to find multiple patterns from one time series and increase forecasting accuracy. Temporal aggregation produces multiple time series and each series has separate properties. We use exponential smoothing methods in the next step to extract various features of time series components in order to forecast time series components for each series. In the final step, we blend predictions of the same kind of components and forecast the target series by the summation of blended predictions. As an empirical example, we forecast traffic accident counts using MAPA and observe that MAPA performance is superior to conventional methods.

• Journal of KIISE:Computer Systems and Theory
• /
• v.28 no.3
• /
• pp.144-154
• /
• 2001

다중 결합 질의에 포함된 다수의 결합 연산지를 효율적으로 처리하기 위해 서는 효율적인 병렬 알고리즘이 필요하다. 최근 다중 해쉬 결합 질의의 처리를 위해 할당 트리를 이용한 방법이 가장 우수한 것으로 알려져 있다. 그러나 이 방법은 실제 결합 시에 할당 트리의 각 노드에서 필연적인 지연이 발생되는 데 이는 튜플-시험 단계에서 외부 릴레이션을 디스크로부터 페이지 단위로 읽는 비용과 이미 읽는 페이지에 대한 해쉬 결합 비용간의 차이에 의해 발생하게 된다. 이들 사이의 실행시간을 가급적 일치시키기 위한 '페이지 실행시간 동기화'기법이 제안되었고 이를 통해 할당 트리 한 노드 실행에 있어서의 지연 시간을 줄일 수 있었다. 하지만 지연 시간을 최소화하기 위해 할당되어질 프로세서의 수 즉, 페이지 실행시간 동기화 계수(k)는 실제 결합 시의 결합률에 따라 상당한 차이를 보이게 되고 결국, 이 차이를 고려하지 않은 다중 해쉬 결합은 성능 면에서 크게 저하될 수밖에 없다. 본 논문에서는 결합 이전에 어느 정도의 결합률을 예측할 수 있다는 전제하에 다중 해쉬 결합 실행 시에 발생할 수 있는 지연 시간을 최소화 할 수 있도록 결합률에 따라 최적의 프로세서들을 노드에 할당함으로서 다중 해쉬 결합의 실행 성능을 개선하였다. 그리고 분석적 비용 모형을 세워 기존 방식과의 다양한 성능 분석을 통해 비용 모형의 타당성을 입증하였다.

• The Korean Journal of Applied Statistics
• /
• v.14 no.1
• /
• pp.91-101
• /
• 2001

It has been shown that the multiple trees predictors are more accurate in reducing test set error than a single tree predictor. There are two ways of generating multiple trees. One is to generate modified training sets by resampling the original training set, and then construct trees. It is known that arcing algorithm is efficient. The other is to perturb randomly the working split at each node from a list of best splits, which is expected to generate reasonably good trees for the original training set. We propose a new combined multiple regression trees predictor which uses the latter multiple regression tree predictor as a predictor based on a modified training set at each stage of arcing. The efficiency of those prediction methods are compared by applying to high throughput screening of chemical compounds for biological effects.

• Journal of Korean Society of Transportation
• /
• v.22 no.5
• /
• pp.123-137
• /
• 2004

The study on traffic assignment is actively being performed which reflect networks status using time. Its background is increasing social needs to use traffic assignment models in not only hardware area of road network plan but also software area of traffic management or control. In addition, multi-class traffic assignment model is receiving study in order to fill a gap between theory and practice of traffic assignment model. This model is made up of two, one of which is multi-driver class and the other multi-vehicle class. The latter is the more realistic because it can be combined with dynamic model. On this background, this study is to build multidynamic model combining the above-mentioned two areas. This has been a theoretic pillar of ITS in which dynamic user equilibrium assignment model is now made an issue, therefore more realistic dynamic model is expected to be built by combining it with multi-class model. In case of multi-vehicle, FIFO would be violated which is necessary to build the dynamic assignment model. This means that it is impossible to build multi-vehicle dynamic model with the existing dynamic assignment modelling method built under the conditions of FIFO. This study builds dynamic network model which could relieve the FIFO conditions. At the same time, simulation method, one of the existing network loading method, is modified to be applied to this study. Also, as a solution(algorithm) area, time dependent shortest path algorithm which has been modified from existing shortest path algorithm and the existing MSA modified algorithm are built. The convergence of the algorithm is examined which is built by calculating dynamic user equilibrium solution adopting the model and algorithm and grid network.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.16 no.2
• /
• pp.329-338
• /
• 2021

This study predicts the amount of wind power generation for rational operation plan of wind power generation and capacity calculation of ESS. For forecasting, we present a method of predicting wind power generation by combining a physical approach and a statistical approach. The factors of wind power generation are analyzed and variables are selected. By collecting historical data of the selected variables, the amount of wind power generation is predicted using deep learning. The model used is a hybrid model that combines a bidirectional long short term memory (LSTM) and a convolution neural network (CNN) algorithm. To compare the prediction performance, this model is compared with the model and the error which consist of the MLP(:Multi Layer Perceptron) algorithm, The results is presented to evaluate the prediction performance.

• Journal of the Institute of Convergence Signal Processing
• /
• v.19 no.4
• /
• pp.199-205
• /
• 2018

So many fast motion estimation algorithms for prediction quality and computational reduction have been published due to tremendous computations of full search algorithm. In the paper, we suggest an algorithm that reduces computation effectively, while keeping prediction quality as almost same as that of the full search. The proposed algorithm based on multiple threshold filter calculates the sum of partial block matching error for each candidate, selects the candidates for the next step, compares the stability of optimal candidates with minimum error, removes impossible candidates, and calculates optimal motion vectors by determining the progress of the next step. By doing that, we can find the minimum error point as soon as possible and obtain the better performance of calculation speed by reducing unnecessary computations. The proposed algorithm can be combined with conventional fast motion estimation algorithms as well as by itself, further reduce computation while keeping the prediction quality as almost same as the algorithms, and prove it in the experimental results.

• KIPS Transactions on Software and Data Engineering
• /
• v.7 no.9
• /
• pp.351-360
• /
• 2018

The LWR (Locally Weighted Regression) model, which is traditionally a lazy learning model, is designed to obtain the solution of the prediction according to the input variable, the query point, and it is a kind of the regression equation in the short interval obtained as a result of the learning that gives a higher weight value closer to the query point. We study on an incremental ensemble learning approach for LWR, a form of lazy learning and memory-based learning. The proposed incremental ensemble learning method of LWR is to sequentially generate and integrate LWR models over time using a genetic algorithm to obtain a solution of a specific query point. The weaknesses of existing LWR models are that multiple LWR models can be generated based on the indicator function and data sample selection, and the quality of the predictions can also vary depending on this model. However, no research has been conducted to solve the problem of selection or combination of multiple LWR models. In this study, after generating the initial LWR model according to the indicator function and the sample data set, we iterate evolution learning process to obtain the proper indicator function and assess the LWR models applied to the other sample data sets to overcome the data set bias. We adopt Eager learning method to generate and store LWR model gradually when data is generated for all sections. In order to obtain a prediction solution at a specific point in time, an LWR model is generated based on newly generated data within a predetermined interval and then combined with existing LWR models in a section using a genetic algorithm. The proposed method shows better results than the method of selecting multiple LWR models using the simple average method. The results of this study are compared with the predicted results using multiple regression analysis by applying the real data such as the amount of traffic per hour in a specific area and hourly sales of a resting place of the highway, etc.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.19 no.5
• /
• pp.713-719
• /
• 2009

In this paper, we develope a water demand forecasting algorithm using AR(Auto-regressive) and MLP(Multi-layer perceptron). To show effectiveness of the proposed method, we analyzed characteristics of time-series data collected in "A" purification plant at Jeon-Buk province during 2007-2008, and then performed the proposed method with various input factors selected through various analyses. As noted in experimental results, the performance of three types model such as multi-regressive, AR(Auto-regressive), and AR+MLP(Auto-regressive + Multi-layer perceptron) show 5.1%, 3.8%, and 3.6% with respect to MAPE(Mean Absolute Percentage Error), respectively. Thus, it is noted that the proposed method can be used to predict short-term water demand for the efficient operation of a water purification plant.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2023.05a
• /
• pp.374-376
• /
• 2023

최근 환율 예측에 관한 다양한 연구가 진행되어 왔다. 이러한 추세에 대응하여 본 연구에서는 Pearson 상관 계수 및 상호 정보를 사용하여 외환 시장의 환율 변동을 분석하는 다중 연결 네트워크를 구축하였다. 본 연구에서는 이러한 구성된 환율 변화에 대한 시공간 의존성 네트워크를 만들고 그래프 기계 학습의 잠재력을 조사하여 예측 정확도를 향상시키려고 노력하였다. 본 연구 결과는 선형 및 비선형 종속 네트워크 모두에 대해 그래프 신경망을 활용한 임베딩을 활용하여 기존의 기계 학습 알고리즘과 결합시킬 경우 환율 변화의 예측력이 향상될 수 있음을 경험적으로 확인하였다. 특히, 이러한 결과는 통화 간 상호 의존성에만 의존하여 추가 데이터 없이 달성되었다. 이 접근 방식은 데이터 효율성을 강화하고 그래프 시각화를 통해 설명력 있는 통찰력을 제공하며 주어진 데이터 세트 내에서 효과적인 데이터를 생성하여 예측력을 높이는 결과로 해석할 수 있다.

• Journal of the Korea Society of Computer and Information
• /
• v.19 no.2
• /
• pp.21-30
• /
• 2014

The use of induction motors has been recently increasing in a variety of industrial sites, and they play a significant role. This has motivated that many researchers have studied on developing fault detection and classification systems of induction motors in order to reduce economical damage caused by their faults. To early identify induction motor faults, this paper effectively estimates spectral envelopes of each induction motor fault by utilizing a linear prediction coding (LPC) analysis technique and an expectation maximization (EM) algorithm. Moreover, this paper classifies induction motor faults into their corresponding categories by calculating Mahalanobis distance using the estimated spectral envelopes and finding the minimum distance. Experimental results show that the proposed approach yields higher classification accuracies than the state-of-the-art conventional approach for both noiseless and noisy environments for identifying the induction motor faults.