• Journal of Information Processing Systems
• /
• v.14 no.2
• /
• pp.418-434
• /
• 2018

With the increasing use of the Internet and electronic documents, automatic text categorization becomes imperative. Several machine learning algorithms have been proposed for text categorization. The k-nearest neighbor algorithm (kNN) is known to be one of the best state of the art classifiers when used for text categorization. However, kNN suffers from limitations such as high computation when classifying new instances. Instance selection techniques have emerged as highly competitive methods to improve kNN through data reduction. However previous works have evaluated those approaches only on structured datasets. In addition, their performance has not been examined over the text categorization domain where the dimensionality and size of the dataset is very high. Motivated by these observations, this paper investigates and analyzes the impact of instance selection on kNN-based text categorization in terms of various aspects such as classification accuracy, classification efficiency, and data reduction.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.12 no.6
• /
• pp.2619-2643
• /
• 2018

In this paper, a new clustering algorithm based on the connected region generation (CRG-clustering) is proposed. It is an effective and robust approach to clustering on the basis of the connectivity of the points and their neighbors. In the new algorithm, a connected region generating (CRG) algorithm is developed to obtain the connected regions and an isolated point set. Each connected region corresponds to a homogeneous cluster and this ensures the separability of an arbitrary data set theoretically. Then, a region expansion strategy and a consensus criterion are used to deal with the points in the isolated point set. Experimental results on the synthetic datasets and the real world datasets show that the proposed algorithm has high performance and is insensitive to noise.

• Proceedings of the Korean Institute of Intelligent Systems Conference
• /
• 2006.11a
• /
• pp.285-288
• /
• 2006

Most previous studies on improving the effectiveness of CBR have focused on the similarity function aspect or optimization of case features and their weights. However, according to some of the prior research, finding the optimal k parameter for the k-nearest neighbor (k-NN) is also crucial for improving the performance of the CBR system. Nonetheless, there have been few attempts to optimize the number of neighbors, especially using artificial intelligence (AI) techniques. In this study, we introduce a genetic algorithm (GA) to optimize the number of neighbors that combine, as well as the weight of each feature. The new model is applied to the real-world case of a major telecommunication company in Korea in order to build the prediction model for the customer profitability level. Experimental results show that our GA-optimized CBR approach outperforms other AI techniques for this mulriclass classification problem.

• Proceedings of the Korea Inteligent Information System Society Conference
• /
• 2005.11a
• /
• pp.588-594
• /
• 2005

사례기반추론(Case-Based Reasoning)은 다양한 예측 문제에 있어서 성공적으로 활용되고 있는 데이터마이닝 기법 중 하나이다. 사례기반추론 시스템의 예측 성능은 예측에 사용되는 최근접이웃(Nearest Neighbor)을 어떻게 설정하느냐에 따라 영향을 받게 된다. 따라서 최근접 이웃을 결정짓는 k 값의 설정은 성공적인 사례기반추론 시스템을 구축하기 위한 중요 요인 중 하나가 된다. 최근접 이웃의 설정에 있어서 대부분의 선행 연구들은 고정된 k 값을 사용하는 사례기반추론 시스템은 k 값을 크게 설정할 경우 최근접 이웃 안에 주어진 오류를 일으킬 수 있으며, k 값이 작게 설정된 경우에는 유사 사례 중 일부만을 예측에 사용하기 때문에 예측 결과의 왜곡을 초래할 수 있다. 본 이웃을 결정함에 있어서 Similarity Threshold를 이용하는 s-NN 방법을 제안하였다. 본 연구의 실험을 위해 UCI(University of california, Irvine) Machine Learning Repository에서 제공하는 두 개의 신용 데이터 셋을 사용하였으며, 실험 결과 s-NN 적용한 CBR 모델이 고정된 k 값을 적용한 전통적인 CBR 모델보다 더 우수한 성능을 보여주었다.

• KSCE Journal of Civil and Environmental Engineering Research
• /
• v.42 no.2
• /
• pp.249-256
• /
• 2022

Various studies are being actively conducted to show that the real-time visualization technology that combines BIM (Building Information Modeling) and AR (Augmented Reality) helps to increase construction management decision-making and processing efficiency. However, when large-capacity BIM data is projected into AR, there are various limitations such as data transmission and connection problems and the image cut-off issue. To improve the high efficiency of visualizing, a mesh optimization algorithm based on the k-nearest neighbors (KNN) classification framework to reconstruct BIM data is proposed in place of existing mesh optimization methods that are complicated and cannot adequately handle meshes with numerous boundaries of the 3D models. In the proposed algorithm, our target BIM model is optimized with the Unity C# code based on triangle centroid concepts and classified using the KNN. As a result, the algorithm can check the number of mesh vertices and triangles before and after optimization of the entire model and each structure. In addition, it is able to optimize the mesh vertices of the original model by approximately 56 % and the triangles by about 42 %. Moreover, compared to the original model, the optimized model shows no visual differences in the model elements and information, meaning that high-performance visualization can be expected when using AR devices.

• Journal of the Korean Data and Information Science Society
• /
• v.17 no.2
• /
• pp.421-436
• /
• 2006

Huang proposed a grey-based nearest neighbor approach to predict accurately missing attribute value in 2004. Our study proposes which way to decide the number of nearest neighbors using not only the deng's grey relational grade but also the wen's grey relational grade. Besides, our study uses not an arithmetic(unweighted) mean but a weighted one. Also, GRG is used by a weighted value when we impute missing values. There are four different methods - DU, DW, WU, WW. The performance of WW(Wen's GRG & weighted mean) method is the best of any other methods. It had been proven by Huang that his method was much better than mean imputation method and multiple imputation method. The performance of our study is far superior to that of Huang.

• Proceedings of the Korean Information Science Society Conference
• /
• 2007.10c
• /
• pp.428-432
• /
• 2007

We present a novel method for determining k nearest neighbors, which accurately recognizes the underlying clusters in a data set. To this end, we introduce the "tiling neighborhood" which is constructed by tiling a number of small local circles rather than a single circle, as existing neighborhood schemes do. Then we formulate the problem of determining the tiling neighborhood as a minimax optimization, leading to an efficient message passing algorithm. For several real data sets, our method outperformed the k-nearest neighbor method. The results suggest that our method can be an alternative to existing for general classification tasks, especially for data sets which have many missing values.

• Proceedings of the Korean Operations and Management Science Society Conference
• /
• 2002.05a
• /
• pp.645-651
• /
• 2002

we propose a k-nearest neighbors(k-NN) based pattern selection method. The method tries to select the patterns that are near the decision boundary and that are correctly labeled. The simulations over synthetic data sets showed promising results: (1) By converting a non-separable problem to a separable one, the search for an optimal error tolerance parameter became unnecessary. (2) SVM training time decreased by two orders of magnitude without any loss of accuracy. (3) The redundant SVM were substantially reduced.

• 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.

• 한국데이터정보과학회:학술대회논문집
• /
• 2006.04a
• /
• pp.55-62
• /
• 2006

Huang proposed a grey-based nearest neighbor approach to predict accurately missing attribute value in 2004. Our study proposes which way to decide the number of nearest neighbors using not only the dong's grey relational grade but also the wen's grey relational grade. Besides, our study uses not an arithmetic(unweighted) mean but a weighted one. Also, GRG is used by a weighted value when we impute a missing values. There are four different methods - DU, DW, WU, WW. The performance of WW(wen's GRG & weighted mean) method is the best of my other methods. It had been proven by Huang that his method was much better than mean imputation method and multiple imputation method. The performance of our study is far superior to that of Huang.