Geraldo Zimbrão

Group and link analysis of multi-relational scientific social networks

Analyzing social networks enables us to detect several inter and intra connections between people in and outside their organizations. We model a multi-relational scientific social network where researchers may have four different types of relationships with each other. We adopt some criteria to enable the modeling of a scientific social network as close as possible to reality. Using clustering techniques with maximum flow measure, we identify the social structure and research communities in a way that allows us to evaluate the knowledge flow in the Brazilian scientific community. Finally, we evaluate the temporal evolution of scientific social networks to suggest/predict new relationships.

Previsão de Estudantes com Risco de Evasão Utilizando Técnicas de Mineração de Dados

A educacao a distancia tem ganhado significativa atencao tanto na academia quanto nas iniciativas governamentais. Neste contexto, cresce tambem a preocupacao com a avaliacao da qualidade dos diversos aspectos destes cursos mediados pelas tecnologias da informacao e comunicacao. Apesar de muitos trabalhos discutirem diversos aspectos da avaliacao em Ead, a literatura carece de relatos de experiencia que, especialmente, abordem os cursos tecnicos a distancia que possuem especificidades relevantes. Assim, este trabalho relata uma experiencia onde se avalia e adapta uma das propostas existentes na literatura de avaliacao mediada por foruns, ao contexto dos cursos tecnicos a distancia, constituindo-se esta adaptacao e sua discussao as principais contribuicoes do mesmo.Em Educacao a Distância mediada por meio de Ambientes Virtuais de Aprendizagem, foruns de discussao sao um instrumento importante e amplamente utilizado na articulacao de debates e discussoes entre os atores envolvidos no processo de ensino e aprendizagem. Com a ampla utilizacao dos foruns muitas mensagens sao trocadas e isso, por vezes, excede a capacidade de monitoramento por parte dos professores e tutores. O presente trabalho apresenta a concepcao de um classificador de mensagens de foruns que classifica as mensagens em positivas ou negativas, a fim de identificar mensagens que necessitam de maior atencao. Este trabalho aplica conceitos de mineracao de textos, com o algoritmo SVM obtendo taxas de acerto satisfatorias.Este artigo apresenta o framework Contagious, cujo proposito e estabelecer diretrizes que norteiem a construcao de redes sociais online orientadas a Difusao de Inovacoes. Compreendo-se o fenomeno das redes sociais online como consequencia natural do carater social do ser humano, vislumbrou-se esse meio tecnologico de comunicacao e interacao social como potencial ferramenta para a extensao de praticas educativas, com vistas a formacao do carater integral do cidadao. Para isso, foi adotada a teoria de Difusao de Inovacoes, propria das ciencias sociais. As contribuicoes deste trabalho, portanto, compreendem duas vertentes: a) o mapeamento de principios de uma teoria social na forma de recursos computacionais e; b) um enfoque orientado a educacao sobre as redes sociais online.A proposta do trabalho consiste em desenvolver um sistema para ser usado no celular como ferramenta de auxilio para alfabetizacao, utilizando-se de imagens e sons como forma de facilitar o aprendizado. Como metodo de desenvolvimento utiliza-se o processo P@PSEduc (Processo Agil para Software Educativo) e a ferramenta JME (Java Micro Edition).O crescente uso e difusao de tecnologias Web, a ubiquidade de ferramentas educacionais vem proporcionado verdadeiras revolucoes nos ambientes de ensino. Atualmente, sabe-se que nao mais se deve tratar alunos de forma homogenea, como se assim os fossem. Em face disso, este artigo apresenta um sistema adaptativo de apoio a aprendizagem colaborativa, cujo tema e a construcao e representacao do conhecimento por meio de mapas mentais multimidia. Tal sistema, baseia-se na Teoria da Carga Cognitiva, cuja preocupacao primaria e a facilidade com a qual as informacoes sao processadas pelos individuos.

Approximate Query Processing in Spatial Databases Using Raster Signatures

A main issue in database area is to process queries efficiently so that the user does not have to wait a long time to get an answer. However, there are many cases where it is not easy to accomplish this requirement. In addition, a fast answer could be more important for the user than receiving an accurate one. In other words, the precision of the query could be lessened, and an approximate answer could be returned, provided that it is much faster than the exact query processing and it has an acceptable accuracy.

Transforming collaborative filtering into supervised learning

We propose an transformation from CF problem to typical supervised learning.The proposed transformation is straightforward and domain-independent.Our transformation greatly outperforms classical collaborative filtering techniques. Collaborative Filtering (CF) is a well-known approach for Recommender Systems (RS). This approach extrapolates rating predictions from ratings given by user on items, which are represented by a user-item matrix filled with a rating r i , j given by an user i on an item j. Therefore, CF has been confined to this data structure relying mostly on adaptations of supervised learning methods to deal with rating predictions and matrix decomposition schemes to complete unfilled positions of the rating matrix. Although there have been proposals to apply Machine Learning (ML) to RS, these works had to transform the rating matrix into the typical Supervised Learning (SL) data set, i.e., a set of pairwise tuples ( x , y ) , where y is the correspondent class (the rating) of the instance x ? R k . So far, the proposed transformations were thoroughly crafted using the domain information. However, in many applications this kind of information can be incomplete, uncertain or stated in ways that are not machine-readable. Even when it is available, its usage can be very complex requiring specialists to craft the transformation. In this context, this work proposes a domain-independent transformation from the rating matrix representation to a supervised learning dataset that enables SL methods to be fully explored in RS. In addition, our transformation is said to be straightforward, in the sense that, it is an automatic process that any lay person can perform requiring no domain specialist. Our experiments have proven that our transformation, combined with SL methods, have greatly outperformed classical CF methods.

Mining and Analyzing Multirelational Social Networks

A Social Network is a social structure consisting of individuals or organizations, usually represented by nodes tied by one or more types of interdependency or relationship. This work focuses on using data mining techniques to identify intra and inter organization groups of people with similar profiles that could have relationships amongst them. For this, we construct a multirelational scientific social network where researchers may have four different types of relations with each other. In this paper, we analyze the scientific scenario in Computing Science in Brazil, assessing how researchers in the best universities and research centres collaborate and relate to each other.

WAVE: an architecture for predicting dropout in undergraduate courses using EDM

Predicting the academic progress of student is an issue faced by many public universities in emerging countries. Although, those institutions stores large amounts of educational data, they fail to recognize the students that are in danger to leave the system. This paper presents a novel architecture that uses EDM techniques to predict and to identify those who are at dropout risk. This approach allows academic managers to monitor the progress of the students in each academic semester, identifying the ones in difficult to fulfill their academic requirements. This paper shows initial experimental results using real world data about of three undergraduate engineering courses of one the largest Brazilian public university. According to the experiments, the classifier Naïve Bayes presented the highest true positive rate for all datasets used in the experiments.

Adaptive Normalization: A novel data normalization approach for non-stationary time series

Data normalization is a fundamental preprocessing step for mining and learning from data. However, finding an appropriated method to deal with time series normalization is not a simple task. This is because most of the traditional normalization methods make assumptions that do not hold for most time series. The first assumption is that all time series are stationary, i.e., their statistical properties, such as mean and standard deviation, do not change over time. The second assumption is that the volatility of the time series is considered uniform. None of the methods currently available in the literature address these issues. This paper proposes a new method for normalizing non-stationary heteroscedastic (with non-uniform volatility) time series. The method, named Adaptive Normalization (AN), was tested together with an Artificial Neural Network (ANN) in three forecast problems. The results were compared to other four traditional normalization methods, and showed AN improves ANN accuracy in both short- and long-term predictions.

Building the data warehouse of frequent itemsets in the DWFIST approach

Some data mining tasks can produce such great amounts of data that we have to cope with a new knowledge management problem. Frequent itemset mining fits in this category. Different approaches were proposed to handle or avoid somehow this problem. All of them have problems and limitations. In particular, most of them need the original data during the analysis phase, which is not feasible for data streams. The DWFIST (Data Warehouse of Frequent ItemSets Tactics) approach aims at providing a powerful environment for the analysis of itemsets and derived patterns, such as association rules, without accessing the original data during the analysis phase. This approach is based on a Data Warehouse of Frequent Itemsets. It provides frequent itemsets in a flexible and efficient way as well as a standardized logical view upon which analytical tools can be developed. This paper presents how such a data warehouse can be built.

Neural networks cartridges for data mining on time series

Neural networks is one of the techniques used for time series analysis. The performance of neural networks is affected by some parameters such as neural network structure and the quality of data preprocessing. These parameters need to be explored in order to obtain an optimal neural network. However, the manual establishment of different neural networks configurations for selecting the best ones may be error-prone and time-consuming. This paper proposes the creation of neural networks cartridges to systematically empower neural network performance by means of data mining activities, which obtain an optimal neural network structure. The experiments conducted in this paper use stock market and exchange rate series, and show that the usage of neural network cartridges can lead to configurations that double the performance of some ad-hoc neural network configuration.

Mining and Analyzing Organizational Social Networks Using Minimum Spanning Tree

This work focuses on using data mining techniques to identify intra and inter organization groups of people with similar profiles and that could have relationships among them. Our clustering method identifies clusters with a link mining-based technique that uses the minimum spanning tree to construct group hierarchies. In this paper we analyze the scientific scenario in Computing Science in Brazil, assessing how researchers in the best universities and research centers collaborate and relate to each other.