• Journal of Korea Multimedia Society
• /
• v.20 no.9
• /
• pp.1567-1573
• /
• 2017

The k-nearest neighbor (k-NN) algorithm is one of the most widely used benchmark algorithm in classification. Nowadays it has been further applied to predict time series. However, one of the main concerns of the algorithm applied on short-term electricity load forecasting is high computational burden. In the paper, we propose an approach of dimension reduction that follows the principles of highlighting the temperature effect on electricity load data series. The results show the proposed approach is able to reduce the dimension of the data around 30%. Moreover, with temperature effect highlighting, the approach will contribute to finding similar days accurately, and then raise forecasting accuracy slightly.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.8 no.11
• /
• pp.3896-3915
• /
• 2014

The Reverse k Nearest Neighbor (RkNN) query is valuable for finding objects influenced by a specific object and is widely used in both scientific and commercial systems. However, the influence level of each object is unknown, information that is critical for some applications (e.g. target marketing). In this paper, we propose a new query type, Ordered Reverse k Nearest Neighbor (ORkNN), and make efforts to adapt it in an on-demand scenario. An Order-k Voronoi diagram based approach is used to answer ORkNN queries. In particular, for different values of k, we pre-construct only one Voronoi diagram. Algorithms on both the server and the clients are presented. We also present experimental results that suggest our proposed algorithms may have practical applications.

• Journal of KIISE:Software and Applications
• /
• v.36 no.4
• /
• pp.273-282
• /
• 2009

Missing data is one of the common problems in building analysis or prediction models using software project data. Missing imputation methods are known to be more effective missing data handling method than deleting methods in small software project data. While K nearest neighbor imputation is a proper missing imputation method in the software project data, it cannot use non-missing information of incomplete project instances. In this paper, we propose an approach to missing data imputation for numerical software project data by combining K nearest neighbor and maximum likelihood estimation; we also extend the average absolute error measure by normalization for accurate evaluation. Our approach overcomes the limitation of K nearest neighbor imputation and outperforms on our real data sets.

• Journal of KIISE:Databases
• /
• v.35 no.5
• /
• pp.447-458
• /
• 2008

In this paper, we address an efficient processing scheme for k-nearest neighbor queries to retrieve k static objects in road network databases. Existing methods cannot expect a query processing speed-up by index structures in road network databases, since it is impossible to build an index by the network distance, which cannot meet the triangular inequality requirement, essential for index creation, but only possible in a totally ordered set. Thus, these previous methods suffer from a serious performance degradation in query processing. Another method using pre-computed network distances also suffers from a serious storage overhead to maintain a huge amount of pre-computed network distances. To solve these performance and storage problems at the same time, this paper proposes a novel approach that creates an index for moving objects by approximating their network distances and efficiently processes k-nearest neighbor queries by means of the approximate index. For this approach, we proposed a systematic way of mapping each moving object on a road network into the corresponding absolute position in the m-dimensional space. To meet the triangular inequality this paper proposes a new notion of average network distance, and uses FastMap to map moving objects to their corresponding points in the m-dimensional space. After then, we present an approximate indexing algorithm to build an R*-tree, a multidimensional index, on the m-dimensional points of moving objects. The proposed scheme presents a query processing algorithm capable of efficiently evaluating k-nearest neighbor queries by finding k-nearest points (i.e., k-nearest moving objects) from the m-dimensional index. Finally, a variety of extensive experiments verifies the performance enhancement of the proposed approach by performing especially for the real-life road network databases.

• International Journal of Fuzzy Logic and Intelligent Systems
• /
• v.11 no.2
• /
• pp.84-88
• /
• 2011

In this paper, a new competition strategy for learning vector quantization is proposed. The simple competitive strategy used for learning vector quantization moves the winning prototype which is the closest to the newly given data pattern. We propose a new learning strategy based on k-nearest neighbor prototypes as the winning prototypes. The selection of several prototypes as the winning prototypes guarantees that the updating process occurs more frequently. The design is illustrated with the aid of numeric examples that provide a detailed insight into the performance of the proposed learning strategy.

• Journal of Multimedia Information System
• /
• v.5 no.2
• /
• pp.67-78
• /
• 2018

This article proposes the modified KNN (K Nearest Neighbor) algorithm which considers the feature similarity and is applied to the word categorization. The texts which are given as features for encoding words into numerical vectors are semantic related entities, rather than independent ones, and the synergy effect between the word categorization and the text categorization is expected by combining both of them with each other. In this research, we define the similarity metric between two vectors, including the feature similarity, modify the KNN algorithm by replacing the exiting similarity metric by the proposed one, and apply it to the word categorization. The proposed KNN is empirically validated as the better approach in categorizing words in news articles and opinions. The significance of this research is to improve the classification performance by utilizing the feature similarities.

• Communications for Statistical Applications and Methods
• /
• v.15 no.6
• /
• pp.977-991
• /
• 2008

This study is about the k-nearest neighbor-based approach for the estimation of mutual information when the type of target variable is categorical and continuous. The results of Monte-Carlo simulation and experiments with real-world data show that k=1 is preferable. In practical application with real world data, our study shows that jittering and bootstrapping is needed.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.9 no.2
• /
• pp.742-762
• /
• 2015

In this study, we propose a distance metric learning approach called discriminant metric learning (DML) for face verification, which addresses a binary-class problem for classifying whether or not two input images are of the same subject. The critical issue for solving this problem is determining the method to be used for measuring the distance between two images. Among various methods, the large margin nearest neighbor (LMNN) method is a state-of-the-art algorithm. However, to compensate the LMNN's entangled data distribution due to high levels of appearance variations in unconstrained environments, DML's goal is to penalize violations of the negative pair distance relationship, i.e., the images with different labels, while being integrated with LMNN to model the distance relation between positive pairs, i.e., the images with the same label. The likelihoods of the input images, estimated using DML and LMNN metrics, are then weighted and combined for further analysis. Additionally, rather than using the k-nearest neighbor (k-NN) classification mechanism, we propose a verification mechanism that measures the correlation of the class label distribution of neighbors to reduce the false negative rate of positive pairs. From the experimental results, we see that DML can modify the relation of negative pairs in the original LMNN space and compensate for LMNN's performance on faces with large variances, such as pose and expression.

• Proceedings of the Korean Association of Geographic Inforamtion Studies Conference
• /
• 2008.10a
• /
• pp.297-302
• /
• 2008

The positioning technology for a moving object is an important and essential component of ubiquitous computing environment and applications, for which Radio Frequency Identification(RFID) has been considered as a core technology. RFID-based positioning system calculates the position of moving object based on k-nearest neighbor(k-nn) algorithm using detected k-tags which have known coordinates and kcan be determined according to the detection range of RFID system. In this paper, RFID-based positioning system determines the position of moving object not using weight factor which depends on received signal strength but assuming that tags within the detection range always operate and have same weight value. Because the latter system is much more economical than the former one. The geometries of tags were determined with considerations in huge buildings like office buildings, shopping malls and warehouses, so they were determined as the line in I-Dimensional space, the square in 2-Dimensional space. In 1-Dimensional space, the optimal detection range is determined as 125% of the tag spacing distance through the analytical and numerical approach. Here, the analytical approach means a mathematical proof and the numerical approach means a simulation using matlab. But the analytical approach is very difficult in 2-Dimensional space, so through the numerical approach, the optimal detection range is determined as 134% of the tag spacing distance in 2-Dimensional space. This result can be used as a fundamental study for designing RFID-based positioning system.

• Journal of Electrical Engineering and Technology
• /
• v.11 no.6
• /
• pp.1872-1879
• /
• 2016

The techniques to recycle and reuse plastics attract public attention. These public attraction and needs result in improving the recycling technique. However, the identification technique for black plastic wastes still have big problem that the spectrum extracted from near infrared radiation spectroscopy is not clear and is contaminated by noise. To overcome this problem, we apply Raman spectroscopy to extract a clear spectrum of plastic material. In addition, to improve the classification ability of fuzzy Radial Basis Function Neural Networks, we apply supervised learning based clustering method instead of unsupervised clustering method. The conditional fuzzy C-Means clustering method, which is a kind of supervised learning based clustering algorithms, is used to determine the location of radial basis functions. The conditional fuzzy C-Means clustering analyzes the data distribution over input space under the supervision of auxiliary information. The auxiliary information is defined by using k Nearest Neighbor approach.