Leandro A. Silva

A SOM combined with KNN for classification task

Classification is a common task that humans perform when making a decision. Techniques of Artificial Neural Networks (ANN) or statistics are used to help in an automatic classification. This work addresses a method based in Self-Organizing Maps ANN (SOM) and K-Nearest Neighbor (KNN) statistical classifier, called SOM-KNN, applied to digits recognition in car plates. While being much faster than more traditional methods, the proposed SOM-KNN keeps competitive classification rates with respect to them. The experiments here presented contrast SOM-KNN with individual classifiers, SOM and KNN, and the results are classification rates of 89.48±5.6, 84.23±5.9 and 91.03±5.1 percent, respectively. The equivalency between SOM-KNN and KNN recognition results are confirmed with ANOVA test, which shows a p-value of 0.27.

A comparison of dimensionality reduction methods using topology preservation indexes

Due to the remarkable technological developments experienced in recent decades, the vast amount of data had created new opportunities and challenges in the field of knowledge discovery and data mining. Factors like size and high dimensionality of databases adds difficulties to the complex task of discovering patterns hidden in masses of data. The feasibility of highdimensional data exploration depends on techniques known as dimensionality reduction methods. When class labels are available, an optimization function can be used to maximize intra class cohesion and inter class separation. However, in many practical situations information about class is not available. This paper focuses on unsupervised dimensionality reduction techniques, an important phase in exploratory data analysis. Six important methods are described: Principal components analysis, Sammon projection, Autoassociative Neural network, Kohonen maps, Isomap and Locally Linear Embedding. Three quality indexes are proposed to try to quantify to some degree the topology preservation between input and output spaces. Comparisons are performed using benchmark data sets. Results and tests focused two-dimensional projections for data visualization purposes.

Prototype Generation Using Self-Organizing Maps for Informativeness-Based Classifier

The k nearest neighbor is one of the most important and simple procedures for data classification task. The kNN, as it is called, requires only two parameters: the number of k and a similarity measure. However, the algorithm has some weaknesses that make it impossible to be used in real problems. Since the algorithm has no model, an exhaustive comparison of the object in classification analysis and all training dataset is necessary. Another weakness is the optimal choice of k parameter when the object analyzed is in an overlap region. To mitigate theses negative aspects, in this work, a hybrid algorithm is proposed which uses the Self-Organizing Maps (SOM) artificial neural network and a classifier that uses similarity measure based on information. Since SOM has the properties of vector quantization, it is used as a Prototype Generation approach to select a reduced training dataset for the classification approach based on the nearest neighbor rule with informativeness measure, named iNN. The SOMiNN combination was exhaustively experimented and the results show that the proposed approach presents important accuracy in databases where the border region does not have the object classes well defined.

Learning Embedded Data Structure with Self-Organizing Maps

Self-Organizing Map (SOM) is undoubtedly one of the most famous and successful artificial neural network approaches. Since the SOM is related with the Vector Quantization learning process, minimizing error quantization and maximizing topology preservation can be concurrent tasks. Besides, even with some metrics, sometimes the analysis of the map results depends on the user and poses an additional difficulty when the user deals with high dimensional data. This work discusses a proposal of relocating the voted map units after the training phase in order to minimize the quantization error and evaluate the impact in the topology preservation. The idea is to enhance the visualization of embedded data structure from input samples using the SOM.

Pattern recognition in mammographic images used by the residents in mammography

This paper presents a computational strategy for content based image retrieval (CBIR-Content-Based Image Retrieval), considering the similarity in relation to an image already selected. The identification of similarity is obtained by feature extraction, using the technique of wavelet combined with Hu moments. The classification of mammographic is performed using Artificial Neural Networks, through the classifier Self-Organizing Map (SOM). The proposed method is tested with a database of the Laboratory of Medical Image Classification (QUALIM) Department of Diagnostic Imaging, Federal University of São Paulo (UNIFESP).

SOMM – self-organized manifold mapping

The Self Organizing Map (SOM) [1] proposed by Kohonen has proved to be remarkable in terms of its range of applications. It can be used for high dimensional space visualization, pattern recognition, input space dimensionality reduction and for generating prototyping to extrapolate information. Basically, tasks conducted by the SOM method are closely related with input space mapping in order to preserve topological and metric relationship between samples. These maps are meant to create a low dimensional output representation of high dimensional input space. Although maps higher than two dimensions can be created by SOM, it is common to work with the limit of one or two dimensions. This work presents a methodology named SOMM (Self-Organized Manifold Mapping) that can be useful to discover structures and clusters of input dataset using the SOM map as a representation of data distribution structure.

Data classification combining Self-Organizing Maps and Informative Nearest Neighbor

The task of classifying data has been addressed in various works, and has been utilized in various areas of application, such as medicine, industry, marketing, financial market and many others. This work will present a data classifier proposal that combines the SOM (Self-Organizing Map) neural network with INN (Informative Nearest Neighbors). The combination of these two algorithms will be called in this work as SOM-INN. The classification will be done in a two steps task: in the first the SOM has a functionality to reduce the dataset through a process that utilizes the winning neuron concept and, in the second step, the dataset objects selected will be utilized as reference for the INN algorithm to decide about the classification utilizing the most informative object of the reduced dataset. Experiments using 14 public datasets will be made comparing classic classification algorithms from the indicators of accuracy and time consumed in the classification process. The results obtained show that the proposed SOM-INN algorithm, when compared with the other classifiers, presents better accuracy rates in databases where the border region does not have the object classes well distributed. However, its main differential is in the classification time, which is extremely important for real applications.

Thresholding the Courtesy Amount of Brazilian Bank Checks Using a Local Methodology

This paper presents a new thresholding methodology for complex background images with an application to the courtesy amount of Brazilian bank checks. Courtesy amount images present a complex background and the proposal of an automatic thresholding process brings benefits to other steps in bank check clearance, such as the Optical Character Recognition (OCR). Experimental results showed that the proposed methodology yields good results, with average accuracy over 95 %, superior to standard methods from the literature.

BIG DATA/ANALYTICS: FORMAÇÃO E GESTÃO DE CIENTISTAS DE DADOS

A analise adequada do grande volume de dados que vem sendo gerado por sistemas convencionais de computador, redes sociais, sensores etc., tende a se tornar fator critico para as organizacoes, pois essa analise pode gerar informacoes fundamentais para o sucesso das mesmas. Ha, no entanto, uma grande carencia de profissionais habilitados a fazer essa analise. Este trabalho discute aspectos ligados as habilidades necessarias a esses profissionais e a sua formacao e gestao, apos apresentar uma visao geral de Big Data e Analytics, que compoem o ambiente onde esses profissionais atuarao. O principal objetivo do trabalho e fornecer subsidios aqueles envolvidos com o assunto.

Fine-tuning of the SOMkNN classifier

Classification is an important data mining task used in decision-making processes. Techniques such as Artificial Neural Networks (ANN) and Statistics are used to help in an automatic classification. In a previous work, we proposed a method for classification problems based on Self-Organizing Maps ANN (SOM) and k Nearest Neighbor (kNN) statistical classifier. The SOMkNN classifier, as we call this combination, is much faster than the traditional kNN and it keeps equivalent rates results. We propose a fine-tuning for this classifier here, which consists of a neuron relocation of the SOM map. The experiments presented compare SOMkNN with and without fine-tuning. Experiments using 8 databases, 6 of which are available in the UCI repository, the fine-tuning results are an improvement classification rate in 7 databases and in the last one the result is the same. The results indicate a trend of classification rate improvement with the application of the fine tuning technique. The gain in rate is approximately 1.2% and experiments were performed in order to correlate the results.