• Management Science and Financial Engineering
• /
• v.16 no.2
• /
• pp.53-65
• /
• 2010

Despite the overall success of neighbor-based CF methods, there are some fundamental questions about neighbor selection and prediction mechanism including arbitrary similarity, over-fitting interpolation weights, no trust consideration between neighbours, etc. This paper proposes a simple method to compute absolute interpolation weights based on similarity values. In order to supplement the method, two schemes are additionally devised for high-quality neighbour selection and trust metrics based on co-ratings. The former requires that one or more neighbour's similarity should be better than a pre-specified level which is higher than the minimum level. The latter gives higher trust to neighbours that have more co-ratings. Experimental results show that the proposed method outperforms the pure IBCF by about 8% improvement. Furthermore, it can be easily combined with other predictors for achieving better prediction quality.

• Journal of Information Technology Applications and Management
• /
• v.13 no.1
• /
• pp.77-94
• /
• 2006

Case-based Reasoning (CBR) is widely used in business and industry prediction. It is suitable to solve complex and unstructured business problems. Recently, the prediction accuracy of CBR has been enhanced by not only various machine learning algorithms such as genetic algorithms, relative weighting of Artificial Neural Network (ANN) input variable but also data mining technique such as feature selection, feature weighting, feature transformation, and instance selection As a result, CBR is even more widely used today in business area. In this study, we investigated the usefulness of the CBR method in forecasting success in implementing ERP systems. We used a CBR method based on the feature weighting technique to compare the performance of three different models : MDA (Multiple Discriminant Analysis), GECBR (GEneral CBR), FWCBR (CBR with Feature Weighting supported by Analytic Hierarchy Process). The study suggests that the FWCBR approach is a promising method for forecasting of successful ERP implementation in Small and Medium sized Enterprises.

• International Journal of Fuzzy Logic and Intelligent Systems
• /
• v.14 no.3
• /
• pp.209-215
• /
• 2014

Sequence tagging is the task of predicting frame-wise labels for a given input sequence and has important applications to diverse domains. Conventional methods such as maximum likelihood (ML) learning matches global features in empirical and model distributions, rather than local features, which directly translates into frame-wise prediction errors. Recent probabilistic sequence models such as conditional random fields (CRFs) have achieved great success in a variety of situations. In this paper, we introduce a novel discriminative CRF learning algorithm to minimize local feature mismatches. Unlike overall data fitting originating from global feature matching in ML learning, our approach reduces the total error over all frames in a sequence. We also provide an efficient gradient-based learning method via gradient forward-backward recursion, which requires the same computational complexity as ML learning. For several real-world sequence tagging problems, we empirically demonstrate that the proposed learning algorithm achieves significantly more accurate prediction performance than standard estimators.

• Journal of the Korea Institute of Military Science and Technology
• /
• v.19 no.3
• /
• pp.319-329
• /
• 2016

Current research trends of prediction of possible structures, synthesis and explosive characteristics of polynitrogen molecules(PNs) are reviewed. Theoretically PNs are composed only of nitrogen atoms, in which N-N bonds are either single or double bonds, and thus when these molecules decompose, release of enormous energy is accompanied. From the middle of 20th century energetic material chemists have been seeking possible structures and the methods of synthesis of these new materials. As a results, from $N_4$ to $N_{60}$ together with their ions are predicted, and experimental chemists have been trying to synthesize these materials with a few success, including the famous ${N_5}^+$ ion in 1999. Although experimental successes are very rare beyond $N_5$ until today, the author believes that renovative ideas together with sincere efforts will bring someday next generation of high energy materials such as nitrogen fullerene($N_{60}$) in reality.

• Proceedings of the Korean Society of Computer Information Conference
• /
• 2020.01a
• /
• pp.15-18
• /
• 2020

특정 영화의 스크린 독과점이 꾸준히 논란이 되고 있다. 본 논문에서는 영화 스크린 분배의 불평등성을 지적하고 이에 대한 개선을 요구할 근거로 머신러닝 기법을 활용한 영화 관람객 예측 모델을 제안한다. 이에 따라 KOBIS, 네이버 영화, 트위터, 구글 트렌드에서 수집한 3,143개의 영화 데이터를 이용하여 랜덤포레스트와 그라디언트 부스팅 기법을 활용한 영화 관람객 예측 모델을 구현하였다. 모델 평가 결과, 그라디언트 부스팅 모델의 RMSE는 600,486, 랜덤포레스트 모델의 RMSE는 518,989로 랜덤포레스트 모델의 예측력이 더 높았다. 예측력이 높았던 랜덤포레스트 모델을 활용, 상영관을 크게 확보하지 못 했던 봉준호 감독의 영화 '옥자'의 상영관 수를 조절하여 관람객 수를 예측, 6,345,011명이라는 결과를 제시한다.

• The Journal of Korea Robotics Society
• /
• v.19 no.2
• /
• pp.196-202
• /
• 2024

The ability to understand given environments and plan a sequence of actions leading to goal state is crucial for personal service robots. With recent advancements in deep learning, numerous studies have proposed methods for state representation in planning. However, previous works lack explicit information about relationships between objects when the state observation is converted to a single visual embedding containing all state information. In this paper, we introduce graph-based state representation that incorporates both object and relationship features. To leverage these advantages in addressing the task planning problem, we propose a Graph Neural Network (GNN)-based subgoal prediction model. This model can extract rich information about object and their interconnected relationships from given state graph. Moreover, a search-based algorithm is integrated with pre-trained subgoal prediction model and state transition module to explore diverse states and find proper sequence of subgoals. The proposed method is trained with synthetic task dataset collected in simulation environment, demonstrating a higher success rate with fewer additional searches compared to baseline methods.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.25 no.10
• /
• pp.1435-1440
• /
• 2021

With the advent of the low interest rate era, many investors are flocking to the stock market. In the past stock market, people invested in stocks labor-intensively through company analysis and their own investment techniques. However, in recent years, stock investment using artificial intelligence and data has been widely used. The success rate of stock prediction through artificial intelligence is currently not high, so various artificial intelligence models are trying to increase the stock prediction rate. In this study, we will look at various artificial intelligence models and examine the pros and cons and prediction rates between each model. This study investigated as stock prediction programs using artificial intelligence artificial neural network (ANN), deep learning or hierarchical learning (DNN), k-nearest neighbor algorithm(k-NN), convolutional neural network (CNN), recurrent neural network (RNN), and LSTMs.

• Atmosphere
• /
• v.32 no.4
• /
• pp.353-365
• /
• 2022

This paper evaluates precipitation forecast skill of Global/Regional Integrated Model system (GRIMs) over South Korea in a boreal winter from December 2013 to February 2014. Three types of precipitation are classified based on development mechanism: 1) convection type (C type), 2) low pressure type (L type), and 3) orographic type (O type), in which their frequencies are 44.4%, 25.0%, and 30.6%, respectively. It appears that the model significantly overestimates precipitation occurrence (0.1 mm d^-1) for all types of winter precipitation. Objective measured skill scores of GRIMs are comparably high for L type and O type. Except for precipitation occurrence, the model shows high predictability for L type precipitation with the most unbiased prediction. It is noted that Equitable Threat Score (ETS) is inappropriate for measuring rare events due to its high dependency on the sample size, as in the case of Critical Success Index as well. The Symmetric Extreme Dependency Score (SEDS) demonstrates less sensitivity on the number of samples. Thus, SEDS is used for the evaluation of prediction skill to supplement the limit of ETS. The evaluation via SEDS shows that the prediction skill score for L type is the highest in the range of 5.0, 10.0 mm d^-1 and the score for O type is the highest in the range of 1.0, 20.0 mm d^-1. C type has the lowest scores in overall range. The difference in precipitation forecast skill by precipitation type can be explained by the spatial distribution and intensity of precipitation in each representative case.

• Journal of Digital Convergence
• /
• v.10 no.9
• /
• pp.319-324
• /
• 2012

Vehicle safety communications is one among the important technologies in order to protect a car accident. For this, many protocols forwarding a safe message have studied to protect a chain-reaction collision when a car accident occurs. most of these protocols assume that the time of generating a safe message is the same as an accident's. If a node predicts some traffic hazard and forwards a safe message, a driver can response some action quickly. So, In this paper, we proposes a traffic hazard prediction algorithm using the communication technique. As a result, we show that the frame reception success rate of using our algorithm to the previous protocol improved about 4~5%.

• Coupled systems mechanics
• /
• v.8 no.6
• /
• pp.523-535
• /
• 2019

Artificial Intelligence (AI) is anticipated to be the future of technology. Hence, AI has been applied in various fields over the years and its applications are expected to grow in number with the passage of time. There has been a growing need for accurate, direct, and quick prediction of geotechnical and foundation engineering models especially since the success of each project relies on numerous amounts of data. In this study, two applications of AI in the field of geotechnical and foundation engineering are presented - spatial interpolation of standard penetration test (SPT) data and prediction of consolidation of clay. SPT and soil profile data may be predicted and estimated at any location and depth at a site that has no available borehole test data using artificial intelligence techniques such as artificial neural networks (ANN) based on available geospatial information from nearby boreholes. ANN can also be used to accelerate the calculation of various theoretical methods such as the one-dimensional consolidation theory of clay with high efficiency by using lesser computation resources. The results of the study showed that ANN can be a valuable, powerful, and practical tool in providing various information that is needed in geotechnical and foundation design.