• ETRI Journal
• /
• v.34 no.6
• /
• pp.922-931
• /
• 2012

With the increasing demands for mobile wireless sensor networks in recent years, designing an energy-efficient clustering and routing protocol has become very important. This paper provides an analytical model to evaluate the power consumption of a mobile sensor node. Based on this, a clustering algorithm is designed to optimize the energy efficiency during cluster head formation. A genetic algorithm technique is employed to find the near-optimal threshold for residual energy below which a node has to give up its role of being the cluster head. This clustering algorithm along with a hybrid routing concept is applied as the near-optimal energy-efficient routing technique to increase the overall efficiency of the network. Compared to the mobile low energy adaptive clustering hierarchy protocol, the simulation studies reveal that the energy-efficient routing technique produces a longer network lifetime and achieves better energy efficiency.

• Proceedings of the KSPS conference
• /
• 2006.05a
• /
• pp.41-44
• /
• 2006

The purpose of state tying is to reduce the number of models and to use relatively reliable output probability distributions. There are two approaches: one is top down clustering and the other is bottom up clustering. For seen data, the performance of bottom up approach is better than that of top down approach. In this paper, we propose a new clustering technique that can enhance the undertrained triphone clustering performance. The basic idea is to tie unreliable triphones before clustering. An unreliable triphone is the one that appears in the training data too infrequently to train the model accurately. We propose to use monophone distance to preprocess these unreliable triphones. It has been shown in a pilot experiment that the proposed method reduces the error rate significantly.

• Journal of Intelligence and Information Systems
• /
• v.9 no.1
• /
• pp.179-193
• /
• 2003

Clustering is a data mining method which help discovering interesting data groups in large databases. In traditional data clustering, similarity between objects in the cluster is measured by pairwise similarity of objects. But we devise an advanced measurement called item similarity in this paper, in terms of nature of clustering transaction data and use this measurement to perform clustering. This new algorithm show the similarity by accepting the concept of relationship between different attributes. With this item similarity measurement, we develop an efficient clustering algorithm for target marketing in each group.

• Journal of Korea Multimedia Society
• /
• v.13 no.10
• /
• pp.1534-1546
• /
• 2010

In sensor network systems, various hierarchical clustering schemes have been proposed in order to efficiently maintain the energy consumption of sensor nodes. Most of these schemes, however, are hardly applicable in practice since these schemes might produce unbalanced clusters or randomly distributed clusters without taking into account of the distribution of sensor nodes. To overcome the limitations of such hierarchical clustering schemes, we propose a novel scheme called CSM(Clustering using Split & Merge algorithm), which exploits node split and merge algorithm of tree-based indexing structures to efficiently construct clusters. Our extensive performance studies show that the CSM constructs highly balanced clustering in a energy efficient way and achieves higher performance up to 1.6 times than the previous clustering schemes, under various operational conditions.

• The Journal of the Acoustical Society of Korea
• /
• v.26 no.5
• /
• pp.199-206
• /
• 2007

In this paper we propose a new clustering algorithm that performs clustering the feature vectors for the speaker identification. Unlike typical clustering approaches, the proposed method performs the clustering without the initial guesses of locations of the cluster centers and a priori information about the number of clusters. Cluster centers are obtained incrementally by adding one cluster center at a time through the boundary subtractive clustering algorithm. The number of clusters is obtained from investigating the mutual relationship between clusters. The experimental results for artificial datum and TIMIT DB show the effectiveness of the proposed algorithm as compared with the conventional methods.

• Korean Journal of Community Nutrition
• /
• v.6 no.1
• /
• pp.97-107
• /
• 2001

The clustering of insulin resistance with hypertension, glucose intolerance, hyperinsulinemia, increased triglyceride and decreased HDL cholesterol levels, and central and overall obesity has been called syndrome X, or the insulin resistance syndrome(IRS). To develop a nutrition service for IRS, this study was performed to evaluate the prevalence of each component of the metabolic abnormalities of IRS and analyze the clustering pattern of IRS among subjects living in the Taegu community. Participants in this study were 9234(mean age ; M/F 48/40yrs);63.5% were men, 24.4% were obese, 13.3% had hypertension. 3.7% had hyperglycemia, and 32.4% had hyperlipidemia. The IRS was defined as the coexistence of two or more components among metabolic abnormalities; obesity, hypertension. hyperglucemia and hyperlipidemia. The prevalence of IRS in Taegu was 19.2%(M/F:20.8%/16.4%), the clustering of these fisk variables was higher in advanced age group. Among the subjects of IRS having two of more diseases, 75.6% were obese, the pattern were similar in men and women. The younger, the higher the prevalence of obesity associated clustering patterns. The prevalence of obesity associated patterns among the hyperglycemia associated clustering patterns was 44.5%. The samples of the representative clustering patterns were obesity and hyperlipidemia (8.0%), hypertension and hyperlipidemia(3.2%), hypertension, obesity and hyperlipiemia(3.1%), hypertension and obesity(2.3%), and hyperglycemia and hyperlipidemia(0.8%). The clustering of obesity and hyperlipidemia until 50 year old groups, and the clustering of hypertension and hyperlipidemia in the 60 and 70 age groups were the most prevalent. We concluded that insulin resistance syndrome was a relatively common disorder in the Taegu community, and prevalence and the characteristics of the intervention strategies for IRS are desired, an effective improvement will be achieved.

• Journal of Information Processing Systems
• /
• v.15 no.2
• /
• pp.359-373
• /
• 2019

For clustering large-scale data, which cannot be loaded into memory entirely, incremental clustering algorithms are very popular. Usually, these algorithms only concern the within-cluster compactness and ignore the between-cluster separation. In this paper, we propose two incremental fuzzy compactness and separation (FCS) clustering algorithms, Single-Pass FCS (SPFCS) and Online FCS (OFCS), based on a fuzzy scatter matrix. Firstly, we introduce two incremental clustering methods called single-pass and online fuzzy C-means algorithms. Then, we combine these two methods separately with the weighted fuzzy C-means algorithm, so that they can be applied to the FCS algorithm. Afterwards, we optimize the within-cluster matrix and betweencluster matrix simultaneously to obtain the minimum within-cluster distance and maximum between-cluster distance. Finally, large-scale datasets can be well clustered within limited memory. We implemented experiments on some artificial datasets and real datasets separately. And experimental results show that, compared with SPFCM and OFCM, our SPFCS and OFCS are more robust to the value of fuzzy index m and noise.

• KIPS Transactions on Computer and Communication Systems
• /
• v.2 no.5
• /
• pp.199-202
• /
• 2013

In the cluster mechanism of the wireless sensor network, the network lifetime is affected by how cluster heads are selected. One of the representative clustering mechanisms, the low-energy adaptive clustering hierarchy (LEACH), selects cluster heads periodically, resulting in high energy consumption in cluster reconstruction. On the other hand, the adaptive clustering algorithm via waiting timer (ACAWT) proposes a non-periodic re-clustering mechanism that reconstructs clusters if the remaining energy level of a cluster head reaches a given threshold. In this paper, we propose a re-clustering mechanism that uses multiple remaining node energy levels and does re-clustering when the remaining energy level of a cluster head reaches one level lower. Also, in determining cluster heads, both of the number of neighbor nodes and the remaining energy level are considered so that cluster heads can be more evenly placed. From the simulations based on the Qualnet simulator, we validate that our proposed mechanism outperforms ACAWT in terms of the network lifetime.

• Journal of Korea Society of Industrial Information Systems
• /
• v.17 no.2
• /
• pp.57-64
• /
• 2012

In this paper, the multi-objective genetic algorithm is proposed for the document clustering which is important in the text mining field. The most important function in the document clustering algorithm is to group the similar documents in a corpus. So far, the k-means clustering and genetic algorithms are much in progress in this field. However, the k-means clustering depends too much on the initial centroid, the genetic algorithm has the disadvantage of coming off in the local optimal value easily according to the fitness function. In this paper, the multi-objective genetic algorithm is applied to the document clustering in order to complement these disadvantages while its accuracy is analyzed and compared to the existing algorithms. In our experimental results, the multi-objective genetic algorithm introduced in this paper shows the accuracy improvement which is superior to the k-means clustering(about 20 %) and the general genetic algorithm (about 17 %) for the document clustering.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.22 no.2
• /
• pp.205-211
• /
• 2012

The retrieved documents have to be transformed into proper data structure for the clustering algorithms of statistics and machine learning. A popular data structure for document clustering is document-term matrix. This matrix has the occurred frequency value of a term in each document. There is a sparsity problem in this matrix because most frequencies of the matrix are 0 values. This problem affects the clustering performance. The sparseness of document-term matrix decreases the performance of clustering result. So, this research uses the factor score by factor analysis to solve the sparsity problem in document clustering. The document-term matrix is transformed to document-factor score matrix using factor scores in this paper. Also, the document-factor score matrix is used as input data for document clustering. To compare the clustering performances between document-term matrix and document-factor score matrix, this research applies two typed matrices to self organizing map (SOM) clustering.