• Journal of Intelligence and Information Systems
• /
• v.9 no.3
• /
• pp.117-135
• /
• 2003

This study is to propose a procedure solving scale problem of traditional collaborative filtering (CF) approach. The CF approach generally uses some similarity measures like correlation coefficient. So, as the user of the Website increases, the complexity of computation increases exponentially. To solve the scale problem, this study suggests a clustering model-based approach using Self-Organizing Map (SOM) and RFM (Recency, Frequency, Momentary) method. SOM clusters users into some user groups. The preference score of each item in a group is computed using RFM method. The items are sorted and stored in their preference score order. If an active user logins in the system, SOM determines a user group according to the user's characteristics. And the system recommends items to the user using the stored information for the group. If the user evaluates the recommended items, the system determines whether it will be updated or not. Experimental results applied to MovieLens dataset show that the proposed method outperforms than the traditional CF method comparatively in the recommendation performance and the computation complexity.

• Korean Journal of Environmental Biology
• /
• v.25 no.2
• /
• pp.149-157
• /
• 2007

This study focused on patterning waterbirds occurrences at the western costal area of the Korean Peninsula in winter and relating the occurrence patterns with their environmental factors. Waterbird communities were monitored at 10 different study areas, and the composition of land cover as environmental factors was estimated at each study area. Overall dabbling ducks were the most abundant with 84% of total individuals, followed by shorebird and diving ducks. Species Anae platyrhynchos was the first dominant species, and Anas formosa was the second one. Self-organizing map (SOM), an unsupervised artificial neural network, was applied for patterning wintering waterbird communities, and identified 6 groups according to the differences of communities compositions. Each group reflected the differences of indicator species as well as their habitats.

• The KIPS Transactions:PartB
• /
• v.13B no.2 s.105
• /
• pp.121-126
• /
• 2006

In this paper, we propose an approach to increase the power of clustering Web images by using high-level semantic features from text information relevant to Web images as well as low-level visual features of image itself. These high-level text features can be obtained from image URLs and file names, page titles, hyperlinks, and surrounding text. As a clustering engine, self-organizing map (SOM) proposed by Kohonen is used. In the SOM-based clustering using high-level text features and low-level visual features, the 200 images from 10 categories are divided in some suitable clusters effectively. For the evaluation of clustering powers, we propose simple but novel measures indicating the degrees of scattering images from the same category, and degrees of accumulation of the same category images. From the experiment results, we find that the high-level text features are more useful in SOM-based Web image clustering.

• Journal of Climate Change Research
• /
• v.9 no.3
• /
• pp.209-221
• /
• 2018

The characteristics of heat wave events in Seoul are analyzed using weather station data from Korea Meteorological Administration (KMA) and European Centre for Medium-Range Weather Forecast (ECMWF) ERA-Interim reanalysis data from 1979 to 2016. Heat waves are defined as events in the upper 10th percentile of the daily maximum temperatures. The associated synoptic weather patterns are then classified into six clusters through Self-Organizing Map (SOM) analysis for sea-level pressure anomalies in East Asia. Cluster 1 shows an anti-cyclonic circulation and weak troughs in southeast and west of Korea, respectively. This synoptic pattern leads to southeasterly winds that advect warm and moist air to the Korean Peninsula. Both clusters 2 and 3 are associated with southerly winds formed by an anti-cyclonic circulation over the east of Korea and cyclonic circulation over the west of Korea. Cluster 4 shows a stagnant weather pattern with weak winds and strong insolation. Clusters 5 and 6 are associated with F?hn wind resulting from an anti-cyclonic circulation in the north of the Korean Peninsula. In terms of long-term variations, event frequencies of clusters 4 and 5 show increasing and decreasing trends, respectively. However, other clusters do not show any long-term trends, indicating that the mechanisms that drive heat wave events in Seoul have remained constant over the last four decades.

• Proceedings of the Korea Water Resources Association Conference
• /
• 2022.05a
• /
• pp.482-482
• /
• 2022

하천의 합류부에서는 수질이 다른 유체가 혼합하여 합류 전과 다른 특성을 보인다. 하천의 합류부에서 수질을 효율적으로 관리하기 위해서는 수질의 공간적인 혼합 특성을 규명하는 것이 중요하다. 합류부에서 수질의 공간적인 혼합 특성을 분석하기 위해 본 연구에서는 토폴로지 데이터 분석(topological data analysis, TDA), 자기 조직화 지도(Self-Organizing Map, SOM), k-평균 알고리즘(K-means clustering algorithm) 세 가지 기법을 이용하였다. 세 가지 기법을 비교하여 어떤 알고리즘이 합류부의 수질 변화 특성을 더 뚜렷하게 나타내는지 분석하였다. 수질 변화 비교 인자들은 pH, chlorophyll, DO, Turbidity 등이 있고, 수질 인자들은 YSI를 활용해 측정하였다. 자료의 측정 지역은 낙동강과 황강이 합류하는 지역이며, 보트에 YSI 장비를 부착하고 횡단하여 측정하였다. 측정한 데이터를 R 프로그램을 통해 세 가지 기법을 적용시켜 수질 변화 비교를 분석한다. 토폴로지 데이터 분석(topological data analysis, TDA)은 거대하고 복잡한 데이터로부터 유의미한 정보를 추출하는 데 사용하고, 자기조직화지도(Self-Organizing Map, SOM) 기법은 차원 축소와 군집화를 동시에 수행한다. k-평균 알고리즘(K-means clustering algorithm) 기법은 주어진 데이터를 k개의 클러스터로 묶는 머신러닝 비지도학습에 속하는 알고리즘이다. 세 가지 방법들의 주목적은 클러스터링이다. 클러스터 분석(Cluster analysis)이란 주어진 데이터들의 특성을 고려해 동일한 성격을 가진 여러 개의 그룹으로 대상을 분류하는 데이터 마이닝의 한 방법이다. 군집화 방법들인 TDA, SOM, K-means를 이용해 합류 지역의 수질 특성들을 클러스터링하여 수질 패턴들을 분석해 하천 수질 오염을 방지할 수 있을 것이다. 본 연구에서는 토폴로지 데이터 분석(topological data analysis, TDA), 자기조직화지도(Self-Organizing Map, SOM), k-평균 알고리즘(K-means clustering algorithm) 세 가지 기법을 이용하여 합류부에서의 수질 특성을 비교하며 어떤 기법이 합류의 특성을 더욱 뚜렷하게 나타내는지 규명했다. 합류의 특성을 군집화 방법을 이용해 알게 된다면, 합류부의 수질 변화 패턴을 다른 합류 지역에서도 적용할 수 있을 것으로 기대된다.

• Journal of the Korean Institute of Telematics and Electronics
• /
• v.26 no.5
• /
• pp.85-91
• /
• 1989

In this paper, we propose a neural net system for speech recognition, which is composed of two neural networks. Firstly the self-supervised BP(Back Propagation) network generates the distributed concept corresponding to the activity pattern in the hidden units. And then the self-organizing neural network forms a concept map which directly displays the similarity relations between concepts. By doing the above, the difficulty in learning the conventional BP network is solved and the weak side of BP falling into a pattern matcher is gone, while the strong point of generating the various internal representations is used. And we have obtained the concept map which is more orderly than the Kohonen's SOFM. The proposed neural net system needs not any special preprocessing and has a self-learning ability.

• The Journal of the Acoustical Society of Korea
• /
• v.14 no.2
• /
• pp.101-112
• /
• 1995

In order to construct a feature map-based phoneme classification system for speech recognition, two procedures are usually required. One is clustering and the other is labeling. In this paper, we present a phoneme classification system based on the Kohonen's Self-Organizing Feature Map (SOFM) for clusterer and labeler. It is known that the SOFM performs self-organizing process by which optimal local topographical mapping of the signal space and yields a reasonably high accuracy in recognition tasks. Consequently, SOFM can effectively be applied to the recognition of phonemes. Besides to improve the performance of the phoneme classification system, we propose the learning algorithm combined with the classical K-mans clustering algorithm in fine-tuning stage. In order to evaluate the performance of the proposed phoneme classification algorithm, we first use totaly 43 phonemes which construct six intra-class feature maps for six different phoneme classes. From the speaker-dependent phoneme classification tests using these six feature maps, we obtain recognition rate of $87.2\%$ and confirm that the proposed algorithm is an efficient method for improvement of recognition performance and convergence speed.

• Journal of the Korean Society of Environmental Restoration Technology
• /
• v.26 no.3
• /
• pp.1-17
• /
• 2023

Korea's natural parks are managed by dividing them into four use districts: nature preservation district, natural environment district, cultural heritage district, and park village district within the park under the goal of 'conservation and sustainable use of natural parks'. However, the use districts divided in this way are designated by reflecting the results derived from the simple drawing overlapping method, and there is a limit in that objective and scientific evidence for this is insufficient. In addition, in Taeanhaean National Park, the case of this study, only a very small area of less than 1% of the nature preservation district is designated, and the natural environment district that serves as a buffer space is designated on an excessively wide scale, making it difficult to efficiently manage the national park. Therefore, the use district is not fulfilling its role. In this study, the purpose of this study was to present a method for analyzing the spatial characteristics of natural parks using environmental indicators and unsupervised learning analysis methods to set the use districts of natural parks. In this study, evaluation indicators that can evaluate the natural and human environments were derived, and the distribution patterns for each indicator were analyzed. Afterwards, by applying Self-Organizing Map (SOM) analysis, one of the unsupervised learning analysis methods, districts with similar characteristics were derived in Taeanhaean National Park, and the characteristics of each district were analyzed. As a result of the study, 7 districts with different characteristics were derived in Taeanhaean National Park, and by examining the contribution of each indicator together, it was possible to reveal that each district had different representative characteristics even though it was an adjacent area. This study evaluated natural parks by comprehensively considering the indicators of the natural and human environments. In addition, the SOM method used in the study is meaningful in that it can provide scientific and objective grounds for the existing zoning and apply it to the management plan.

• KSCE Journal of Civil and Environmental Engineering Research
• /
• v.26 no.4B
• /
• pp.389-398
• /
• 2006

The present study compositely applied the self-organizing map (SOM), which is a kind of artificial neural networks (ANNs), and the back propagation algorithm (BPA) for the rainfall-runoff prediction model taking account of the irregular variation of the spatiotemporal distribution of rainfall. To solve the problems from the previous studies on ANNs, such as the overestimation of low flow during the dry season, the underestimation of runoff during the flood season and the persistence phenomenon, in which the predicted values continuously represent the preceding runoffs, we introduced SOM theory for the preprocessing in the prediction model. The theory is known that it has the pattern classification ability. The method proposed in the present research initially includes the classification of the rainfall-runoff relationship using SOM and the construction of the respective models according to the classification by SOM. The individually constructed models used the data corresponding to the respectively classified patterns for the runoff prediction. Consequently, the method proposed in the present study resulted in the better prediction ability of runoff than that of the past research using the usual application of ANNs and, in addition, there were no such problems of the under/over-estimation of runoff and the persistence.

• KSCE Journal of Civil and Environmental Engineering Research
• /
• v.26 no.1B
• /
• pp.61-67
• /
• 2006

Various methods have been applied for the research to model the relationship between rainfall-runoff, which shows a strong nonlinearity. In particular, most researches to model the relationship between rainfall-runoff using artificial neural networks have used back propagation algorithm (BPA), Levenberg Marquardt (LV) and radial basis function (RBF). and They have been proved to be superior in representing the relationship between input and output showing strong nonlinearity and to be highly adaptable to rapid or significant changes in data. The theory of artificial neural networks is utilized not only for prediction but also for classifying the patterns of data and analyzing the characteristics of the patterns. Thus, the present study applied self?organizing map (SOM) based on Kohonen's network theory in order to classify the patterns of rainfall-runoff process and analyze the patterns. The results from the method proposed in the present study revealed that the method could classify the patterns of rainfall in consideration of irregular changes of temporal and spatial distribution of rainfall. In addition, according to the results from the analysis the patterns between rainfall-runoff, seven patterns of rainfall-runoff relationship with strong nonlinearity were identified by SOM.