Search

Density-based clustering has received increasing attention for its ability to handle clusters of arbitrary shapes. However, it still has difficulties in mining clusters of arbitrary densities, especially the clusters of sparse regions in the presence of dense regions. To address this problem, this paper presents a new concept called density decreased chain on the mutual k-NN graph. It starts with the local density center whose density is the highest in the data points connected to this center. Based on the density decreased chain, the concept of the core point is redefined. The density of the core point is close to that of the local density center on the same density decreased chain as the core point. According to its definition, the core point in the data with arbitrary densities can be well identified because the local density centers exist in both sparse and dense regions. Further, intra-cluster density decreased chain is defined to mine subclusters in the core points. After forming the subclusters, the remaining data point is hierarchically assigned to one of these subclusters by the density decreased chains containing this remaining data point. The experiments illustrate the effectiveness of the proposed method. [ABSTRACT FROM AUTHOR]

Although data mining problems require a flat mining table as input, in many real-world applications analysts are interested in finding patterns in a relational database. To this end, new methods and software have been recently developed that automatically add attributes (or features) to a target table of a relational database which summarize information from all other tables. When attributes are automatically constructed by these methods, selecting the important attributes is particularly difficult, because a large number of the attributes are highly correlated. In this setting, attribute selection techniques such as the Least Absolute Shrinkage and Selection Operator (lasso), elastic net, and other machine learning methods tend to under-perform. In this paper, we introduce a novel attribute selection procedure, where after an initial screening step, we cluster the attributes into different groups and apply the group lasso to select both the true attributes groups and then the true attributes. The procedure is particularly suited to high dimensional data sets where the attributes are highly correlated. We test our procedure on several simulated data sets and a real-world data set from a marketing database. The results show that our proposed procedure obtains a higher predictive performance while selecting a much smaller set of attributes when compared to other state-of-the-art methods. [ABSTRACT FROM AUTHOR]

Nonnegative matrix factorization and its graph regularized extensions have received significant attention in machine learning and data mining. However, existing approaches are sensitive to outliers and noise due to the utilization of the squared loss function in measuring the quality of graph regularization and data reconstruction. In this paper, we present a novel robust graph regularized NMF model (RGNMF) to approximate the data matrix for clustering. Our assumption is that there may exist some entries of the data corrupted arbitrarily, but the corruption is sparse. To address this problem, an error matrix is introduced to capture the sparse corruption. With this sparse outlier matrix, a robust factorization result could be obtained since a much cleaned data could be reconstructed. Moreover, the ℓ1-norm function is used to alleviate the influence of unreliable regularization which is incurred by unexpected graphs. That is, the sparse error matrix alleviates the impact of noise and outliers, and the ℓ1-norm function leads to a faithful regularization since the influence of the unreliable regularization errors can be reduced. Thus, RGNMF is robust to unreliable graphs and noisy data. In order to solve the optimization problem of our method, an iterative updating algorithm is proposed and its convergence is also guaranteed theoretically. Experimental results show that the proposed method consistently outperforms many state-of-the-art methods. [ABSTRACT FROM AUTHOR]

Machine Learning techniques have been useful in almost every field of concern. Data Mining, a branch of Machine Learning is one of the most extensively used techniques. The ever-increasing demands in the field of medicine are being addressed by computational approaches in which Big Data analysis, image processing and data mining are on top priority. These techniques have been exploited in the domain of ophthalmology for better retinal fundus image analysis. Blood vessels, one of the most significant retinal anatomical structures are analysed for diagnosis of many diseases like retinopathy, occlusion and many other vision threatening diseases. Vessel segmentation can also be a pre-processing step for segmentation of other retinal structures like optic disc, fovea, microneurysms, etc. In this paper, blood vessel segmentation is attempted through image processing and data mining techniques. The retinal blood vessels were segmented through color space conversion and color channel extraction, image pre-processing, Gabor filtering, image postprocessing, feature construction through application of principal component analysis, k-means clustering and first level classification using Naïve-Bayes classification algorithm and second level classification using C4.5 enhanced with bagging techniques. Association of every pixel against the feature vector necessitates Big Data analysis. The proposed methodology was evaluated on a publicly available database, STARE. The results reported 95.05% accuracy on entire dataset; however the accuracy was 95.20% on normal images and 94.89% on pathological images. A comparison of these results with the existing methodologies is also reported. This methodology can help ophthalmologists in better and faster analysis and hence early treatment to the patients. [ABSTRACT FROM AUTHOR]

The k-means algorithm is widely used for clustering, compressing, and summarizing vector data. We present a fast and memory-efficient GPU-based algorithm for exact k-means, Asynchronous Selective Batched K-means (ASB K-means). Unlike most GPU-based k-means algorithms that require loading the whole dataset onto the GPU for clustering, the amount of GPU memory required to run our algorithm can be chosen to be much smaller than the size of the whole dataset. Thus, our algorithm can cluster datasets whose size exceeds the available GPU memory. The algorithm works in a batched fashion and applies the triangle inequality in each k-means iteration to omit a data point if its membership assignment, i.e., the cluster it belongs to, remains unchanged, thus significantly reducing the number of data points that need to be transferred between the CPU's RAM and the GPU's global memory and enabling the algorithm to very efficiently process large datasets. Our algorithm can be substantially faster than a GPU-based implementation of standard k-means even in situations when application of the standard algorithm is feasible because the whole dataset fits into GPU memory. Experiments show that ASB K-means can run up to 15x times faster than a standard GPU-based implementation of k-means, and it also outperforms the GPU-based k-means implementation in NVIDIA's open-source RAPIDS machine learning library on all the datasets used in our experiments. [ABSTRACT FROM AUTHOR]

Text document classification and clustering is an important learning task which fits to both data mining and machine learning areas. The learning task throws several challenges when it is required to process high dimensional text documents. Word distribution in text documents plays a very key role in learning process. Research related to high dimensional text document classification and clustering is usually limited to application of traditional distance functions and most of the research contributions in the existing literature did not consider the word distribution in documents. In this research, we propose a novel similarity function for feature pattern clustering and high dimensional text classification. The similarity function proposed is used to carry supervised learning based dimensionality reduction. The important feature of this work is that the word distribution before and after dimensionality reduction is the same. Experiment results prove the proposed approach achieves dimensionality reduction, retains the word distribution and obtained better classification accuracies compared to other measures. [ABSTRACT FROM AUTHOR]

We consider the problem of clustering a set of points so as to minimize the maximum intra-cluster dissimilarity, which is strongly NP-hard. Exact algorithms for this problem can handle datasets containing up to a few thousand observations, largely insufficient for the nowadays needs. The most popular heuristic for this problem, the complete-linkage hierarchical algorithm, provides feasible solutions that are usually far from optimal. We introduce a sampling-based exact algorithm aimed at solving large-sized datasets. The algorithm alternates between the solution of an exact procedure on a small sample of points, and a heuristic procedure to prove the optimality of the current solution. Our computational experience shows that our algorithm is capable of solving to optimality problems containing more than 500,000 observations within moderate time limits, this is two orders of magnitude larger than the limits of previous exact methods. [ABSTRACT FROM AUTHOR]