• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제17권3호
• /
• pp.721-739
• /
• 2023

Rating prediction is an important issue in recommender systems, and its accuracy affects the experience of the user and the revenue of the company. Traditional recommender systems use Factorization Machinesfor rating predictions and each feature is selected with the same weight. Thus, there are problems with inaccurate ratings and limited data representation. This study proposes a deep recommendation model based on self-attention Factorization (SAFMR) to solve these problems. This model uses Convolutional Neural Networks to extract features from user and item reviews. The obtained features are fed into self-attention mechanism Factorization Machines, where the self-attention network automatically learns the dependencies of the features and distinguishes the weights of the different features, thereby reducing the prediction error. The model was experimentally evaluated using six classes of dataset. We compared MSE, NDCG and time for several real datasets. The experiment demonstrated that the SAFMR model achieved excellent rating prediction results and recommendation correlations, thereby verifying the effectiveness of the model.

• 한국지능시스템학회논문지
• /
• 제15권4호
• /
• pp.412-417
• /
• 2005

오늘날 인터넷을 기반으로 하는 원격 제어 기술이 급속히 발달하고 있다. 그러나 이러한 원거리 네트워크 기반 제어는 데이터를 전송함에 있어서 지연이 불가피하며, 또한 이 지연이 일정하지 않은 문제점을 지니고 있다. 이러한 네트워크 지연은 시스템의 안정성이나 정확도에 영향을 미친다. 본 논문에서는 네트워크상의 데이터 전송 지연을 고려한 이동 로봇의 원격 운용을 위해 TSK (Takagi-Sugeno-Kang) 퍼지 시스템을 이용하여 전송 지연의 확률 분포 함수와 네트워크 모델을 구하고 이를 전송 지연 예측 알고리즘에 적용하였다. 그리고 컴퓨터 시뮬레이션으로부터 제안된 알고리즘의 실효성을 검증하고, 기존의 예측 알고리즘과의 비교분석을 통하여 그 성능을 평가하였다.

• Nuclear Engineering and Technology
• /
• 제56권9호
• /
• pp.3740-3749
• /
• 2024

To reduce the losses caused by aging failure of insulation gate bipolar transistor (IGBT), which is the core components of nuclear power plant rod position indicating and rod control (RPC) system. It is necessary to conduct studies on its life prediction. The selection of IGBT failure characteristic parameters in existing research relies heavily on failure principles and expert experience. Moreover, the analysis and learning of time-domain degradation data have not been fully conducted, resulting in low prediction efficiency as the monotonicity, time correlation, and poor anti-interference ability of extracted degradation features. This paper utilizes the advantages of the stacked denoising autoencoder(SDAE) network in adaptive feature extraction and denoising capabilities to perform adaptive feature extraction on IGBT time-domain degradation data; establishes a long-short-term memory (LSTM) prediction model, and optimizes the learning rate, number of nodes in the hidden layer, and number of hidden layers using the Gray Wolf Optimization (GWO) algorithm; conducts verification experiments on the IGBT accelerated aging dataset provided by NASA PCoE Research Center, and selects performance evaluation indicators to compare and analyze the prediction results of the SDAE-LSTM model, PSOLSTM model, and BP model. The results show that the SDAE-LSTM model can achieve more accurate and stable IGBT life prediction.

• 한국소성가공학회:학술대회논문집
• /
• 한국소성가공학회 2004년도 제5회 압연심포지엄 신 시장 개척을 위한 압연기술
• /
• pp.175-183
• /
• 2004

In hot strip rolling, a capability for precisely predicting roll force is crucial for sound process control. In the past, on-line prediction models have been developed mostly on the basis of Orowan's theory and its variation. However, the range of process conditions in which desired prediction accuracy could be achieved was rather limited, mainly due to many simplifying assumptions inherent to Orowan's theory. As far as the prediction accuracy is concerned, a rigorously formulated finite element(FE) process model is perhaps the best choice. However, a FE process model in general requires a large CPU time, rendering itself inadequate for on-line purpose. In this report, we present a FE-based on-line prediction model applicable to precision process control in a finishing mill(FM). Described was an integrated FE process model capable of revealing the detailed aspects of the thermo-mechanical behavior of the roll-strip system. Using the FE process model, a series of process simulation was conducted to investigate the effect of diverse process variables on some selected non-dimensional parameters characterizing the thermo-mechanical behavior of the strip. Then, it was shown that an on-line model for the prediction of roll force could be derived on the basis of these parameters. The prediction accuracy of the proposed model was examined through comparison with measurements from the hot strip mill.

• 한국자동차공학회논문집
• /
• 제10권5호
• /
• pp.206-213
• /
• 2002

This paper proposes a vehicle trajectory prediction method for application to vehicle-to-vehicle distance control. This method is based on 2-dimensional kinematics and a Kalman filter has been used to estimate acceleration of the object vehicle. The simulation results using the proposed control method show that the relative distance characteristics can be improved via the trajectory prediction method compared to the customary vehicle stop and go cruise control systems which makes the vehicle remain at a safe distance from a preceding vehicle according to the driver's preference, automatically slow down and come to a full stop behind a preceding vehicle.

• 한국ITS학회 논문지
• /
• 제15권4호
• /
• pp.68-80
• /
• 2016

자율주행 차량은 다양한 센서를 활용하여 사람과 유사한 수준으로 실시간 도로환경 변화를 인지, 환경 변화에 대한 적절한 판단 및 제어를 수행하여야 한다. 특히 영상센서는 차선인식 기능을 통해 주행방향 결정 및 차로이탈 방지 등 조향제어 수행을 위한 인지에 활용된다. 하지만 관련 성능기준은 ADAS(Advanced Driver Assistance System)와 연계된 '운전자 보조' 역할에 초점이 맞춰져, 자율주행시 요구되는 '주체적 상황 인지'를 위한 성능조건과 다를 것으로 판단된다. 본 연구에서는 자율주행시 차선인식 기능이 정상적으로 작동되지 않는 상황이 지속될 때 차량 진행방향과 도로 선형방향의 불일치에 따라 발생되는 횡방향 차로이탈을 차량의 이동 궤적을 기반하여 추정하고, 안전성 확보를 위한 차로이탈 허용 수준 및 영상센서 성능수준을 제시하였다. 분석 결과 승용차 조건에서 차선인식 기능이 1초 이상 연속적인 오작동을 일으킨다면 차로이탈에 의한 위험한 상황에 놓일 수 있는 것으로 나타났다. 따라서 자율주행 차량을 위한 차선인식 기능 평가 시 현재 기준보다 큰 횡방향 차로이탈상황에 대한 검토가 필요할 것으로 판단된다.

• 한국시뮬레이션학회논문지
• /
• 제14권3호
• /
• pp.91-100
• /
• 2005

Grid computing is widely applicable to various fields of industry including process control and manufacturing, military command and control, transportation management, and so on. In a viewpoint of application area, grid computing can be classified to three aspects that are computational grid, data grid and access grid. This paper focuses on computational grid which handles complex and large-scale computing problems. Computational grid is characterized by system dynamics which handles a variety of processors and jobs on continuous time. To solve problems of system complexity and reliability due to complex system dynamics, computational grid needs scheduling policies that allocate various jobs to proper processors and decide processing orders of allocated jobs. This paper proposes a service prediction-based job scheduling model and present its scheduling algorithm that is applicable for computational grid. The service prediction-based job scheduling model can minimize overall system execution time since the model predicts the next processing time of each processing component and distributes a job to a processing component with minimum processing time. This paper implements the job scheduling model on the DEVS modeling and simulation environment and evaluates its efficiency and reliability. Empirical results, which are compared to conventional scheduling policies, show the usefulness of service prediction-based job scheduling.

• 제어로봇시스템학회논문지
• /
• 제13권4호
• /
• pp.296-303
• /
• 2007

This paper proposes a new width control system composed of an ANWC(Automatic Neural network based Width Control) and a fuzzy-PID controller in hot strip finishing mills which aims at obtaining the desirable width. The ANWC is designed using a neural network based width prediction model to minimize a width variation between the measured width and its target value. Input variables for the neural network model are chosen by using the hypothesis testing. The fuzzy-PlD control system is also designed to obtain the fast looper response and the high width control precision in the finishing mill. It is shown through the field test of the Pohang no. 1 hot strip mill of POSCO that the performance of the width margin is considerably improved by the proposed control schemes.

• 제어로봇시스템학회논문지
• /
• 제18권10호
• /
• pp.940-946
• /
• 2012

This paper proposes a method to predict maximum traction for unmanned robots on rough terrain in order to improve traversability. For a traction prediction, we use a friction-slip model based on modified Brixius model derived empirically in terramechanics which is a function of mobility number $B_n$ and slip ratio S. A friction-slip model includes characteristics of various rough terrains where robots are operated such as soil, sandy soil and grass-covered soil. Using a friction-slip model, we build a prediction model for terrain parameters on which we can know maximum static friction and optimal slip with respect to mobility number $B_n$. In this paper, Mobility number $B_n$ is estimated by modified Willoughby Sinkage model which is a function of sinkage z and slip ratio S. Therefore, if sinkage z and slip ratio are measured once by sensors such as a laser sensor and a velocity sensor, then mobility number $B_n$ is estimated and maximum traction is predicted through a prediction model for terrain parameters. Estimation results for maximum traction are shown on simulation using MATLAB. Prediction Performance for maximum traction of various terrains is evaluated as high accuracy by analyzing estimation errors.

• 한국공간구조학회논문집
• /
• 제20권2호
• /
• pp.51-58
• /
• 2020

Recently, deep learning that is the most popular and effective class of machine learning algorithms is widely applied to various industrial areas. A number of research on various topics about structural engineering was performed by using artificial neural networks, such as structural design optimization, vibration control and system identification etc. When nonlinear semi-active structural control devices are applied to building structure, a lot of computational effort is required to predict dynamic structural responses of finite element method (FEM) model for development of control algorithm. To solve this problem, an artificial neural network model was developed in this study. Among various deep learning algorithms, a recurrent neural network (RNN) was used to make the time history response prediction model. An RNN can retain state from one iteration to the next by using its own output as input for the next step. An eleven-story building structure with semi-active tuned mass damper (TMD) was used as an example structure. The semi-active TMD was composed of magnetorheological damper. Five historical earthquakes and five artificial ground motions were used as ground excitations for training of an RNN model. Another artificial ground motion that was not used for training was used for verification of the developed RNN model. Parametric studies on various hyper-parameters including number of hidden layers, sequence length, number of LSTM cells, etc. After appropriate training iteration of the RNN model with proper hyper-parameters, the RNN model for prediction of seismic responses of the building structure with semi-active TMD was developed. The developed RNN model can effectively provide very accurate seismic responses compared to the FEM model.