Search

Emerging Pattern classifiers are accurate and easy to understand classifiers. However, they have two characteristics that can degrade their accuracy: global discretization of numerical attributes and high sensitivity to the support threshold value. In this paper, we introduce a novel algorithm to find emerging patterns without global discretization. Additionally, we propose a new method for building cascades of emerging pattern classifiers, which combines the higher accuracy of classifying with higher support thresholds with the lower levels of abstention of classifying with lower thresholds. Experimental results show that our cascade attains higher accuracy than other state-of-the-art classifiers, including one of the most accurate emerging pattern based classifier. [ABSTRACT FROM AUTHOR]

In this paper, we introduce a new hybrid filter-wrapper method for supervised feature selection, based on the Laplacian Score ranking combined with a wrapper strategy. We propose to rank features with the Laplacian Score to reduce the search space, and then we use this order to find the best feature subset. We compare our method against other based on ranking feature selection methods, namely, Information Gain Attribute Ranking, Relief, Correlation-based Feature Selection, and additionally we include in our comparison a Wrapper Subset Evaluation method. Empirical results over ten real-world datasets from the UCI repository show that our hybrid method is competitive and outperforms in most of the cases to the other feature selection methods used in our experiments. [ABSTRACT FROM AUTHOR]

Ultraviolet Spectra (UVS) analysis is a frequent tool in tasks like diseases diagnosis, drugs detection and hyperspectral remote sensing. A key point in these applications is the UVS comparison function. Although there are several UVS comparisons functions, creating good dissimilarity functions is still a challenge because there are different substances with very similar spectra and the same substance may produce different spectra. In this paper, we introduce a new spectral dissimilarity measure for substances identification, based on the way experts visually match the spectra shapes. We also combine the new measure with the Spectral Correlation Measure. A set of experiments conducted with a database of real substances reveals superior results of the combined dissimilarity, with respect to state-of-the-art measures. We use Receiver Operating Characteristic curve analysis to show that our proposal get the best tradeoff between false positive rates and true positive rates. [ABSTRACT FROM AUTHOR]

Performance and accuracy of a neural network are strongly related to its design. Designing a neural network involves topology (number of neurons, number of layers, number of synapses between layers, etc.), training synapse weights, and parameter selection. Radial basis function neural networks (RBFNNs) could additionally require some other parameters, for example, the means and standard deviations if the activation function of neurons in the hidden layer is a Gaussian function. Commonly, Genetic Algorithms and Evolution Strategies have been used for automatically designing RBFNNs In this work, the use of prototype selection methods for designing a RBFNN is proposed and studied. Experimental results show the viability of designing RBFNNs using prototype selection. [ABSTRACT FROM AUTHOR]

The Nearest Neighbor classifier is a simple but powerful non-parametric technique for supervised classification. However, it is very sensitive to noise and outliers, which could decrease the classifier accuracy. To overcome this problem, we propose two new editing methods based on maximum similarity graphs. Numerical experiments in several databases show the high quality performance of our methods according to classifier accuracy. [ABSTRACT FROM AUTHOR]

Finding a minimal subset of objects that correctly classify the training set for the nearest neighbors classifier has been an active research area in Pattern Recognition and Machine Learning communities for decades. Although finding the Minimal Consistent Subset is not feasible in many real applications, several authors have proposed methods to find small consistent subsets. In this paper, we introduce a novel algorithm for this task, based on support graphs. Experiments over a wide range of repository databases show that our algorithm finds consistent subsets with lower cardinality than traditional methods. [ABSTRACT FROM AUTHOR]

Frequent connected subgraph mining (FCSM) is an interesting task with wide applications in real life. Most of the previous studies are focused on pruning search subspaces or optimizing the subgraph isomorphism (SI) tests. In this paper, a new property to remove all duplicate candidates in FCSM during the enumeration is introduced. Based on this property, a new FCSM algorithm called gdFil is proposed. In our proposal, the candidate space does not contain duplicates; therefore, we can use a fast evaluation strategy for reducing the cost of SI tests without wasting memory resources. Thus, we introduce a data structure to reduce the cost of SI tests. The performance of our algorithm is compared against other reported algorithms. [ABSTRACT FROM AUTHOR]

In this work, a new method for class-specific feature selection, which selects a possible different feature subset for each class of a supervised classification problem, is proposed. Since conventional classifiers do not allow using a different feature subset for each class, the use of a classifier ensemble and a new decision rule for classifying new instances are also proposed. Experimental results over different databases show that, using the proposed method, better accuracies than using traditional feature selection methods, are achieved. [ABSTRACT FROM AUTHOR]

In this paper, a new algorithm for mining frequent connected subgraphs called gRed (graph Candidate Reduction Miner) is presented. This algorithm is based on the gSpan algorithm proposed by Yan and Jan. In this method, the mining process is optimized introducing new heuristics to reduce the number of candidates. The performance of gRed is compared against two of the most popular and efficient algorithms available in the literature (gSpan and Gaston). The experimentation on real world databases shows the performance of our proposal overcoming gSpan, and achieving better performance than Gaston for low minimal support when databases are large. [ABSTRACT FROM AUTHOR]

The edition process is an important task in supervised classification because it helps to reduce the size of the training sample. On the other hand, Instance-Based classifiers store all the training set indiscriminately, which in almost all times, contains useless or harmful objects, for the classification process. Therefore it is important to delete unnecessary objects to increase both classification speed and accuracy. In this paper, we propose an edition method based on sequential search and we present an empirical comparison between it and some other decremental edition methods. [ABSTRACT FROM AUTHOR]