Brazil

Unbalanced community structures is a common phenomenon when representing real-world systems using Complex Networks; however, not all community detection model can correctly identify communities of different sizes. In this paper, we propose a community detection technique focusing on the detection of unbalanced communities. The method augments the Particle Competition model: a bio-inspired technique that employs a set of particles (random walkers) into the network to compete for its nodes. However, the detection of unbalanced communities in the original Particle Competition is impossible because every particle shares the same ability to defend and attack nodes. The proposed model introduces a second stage into the system, named Regularization, which explicitly handles the detection of unbalanced communities. This mechanism uses the neighborhood of the nodes to create a custom preference guide for each particle, specifically tailored to the community of the particle. As a result, the model can precisely detect unbalanced community structures. Furthermore, the model achieves higher accuracy and faster computational speed compared to the original Particle Competition model.

BRACIS 2020

Particle Competition for Unbalanced Community Detection in Complex Networks

particle competition

unbalanced community detection

complex networks

The 9th Brazilian Conference on Intelligent Systems (BRACIS), the 16h Encontro Nacional de Inteligência Artificial e Computacional (ENIAC), and the 12th Concurso de Teses e Dissertações em Inteligência Artificial e Computacional (CTDIAC) was held from 20 to 23 October, 2020.

BRACIS originated from the combination of the two most important scientific events in Brazil in Artificial Intelligence (AI) and Computational Intelligence (CI): the Brazilian Symposium on Artificial Intelligence - SBIA (22 editions), and the Brazilian Symposium on Neural Networks - SBRN (13 editions). ENIAC is a national forum for researchers, practitioners, educators and students to present and discuss innovations, trends, experiences, developments and work in progress in the fields of Artificial and Computational Intelligence. CTDIAC is a satellite event whose goals are to disseminate and award the best dissertations and theses (MSc and PhDs concluded in Brazilian universities from June 2018 to May 2020) in Artificial and Computational Intelligence areas.

In this year, the Symposium on Knowledge Discovery, Mining and Learning (KDMiLe) will be held in conjunction with BRACIS 2020. This sysposium aims at integrating researchers, practitioners, developers, students and users to present their research results, to discuss ideas, and to exchange techniques, tools, and practical experiences – related to the Data Mining and Machine Learning areas.

technical paper

One of the main challenges in Machine Learning and Data Mining fields is the treatment of large Data Streams in the presence of Concept Drifts. This paper presents two families of ensemble algorithms designed to adapt to abrupt and gradual concept drifts. The families Fast Stacking of Ensembles boosting the Old (FASEO) and Fast Stacking of Ensembles boosting the Best (FASEB) are adaptations of the Fast Adaptive Stacking of Ensembles (FASE) algorithm to improve run-time, without presenting a significant decrease in terms of accuracy when compared to the original FASE. In order to achieve a more efficient model, adjustments were made in the update strategy and voting procedure of the ensemble. To evaluate the methods, Naive Bayes (NB) and Hoeffding Tree (HT) are used, as learners, to compare the performance of the algorithms on artificial and real-world data-sets. An experimental investigation with a total of 32 experiments and the application of Friedman and Bonferroni-Dunn statistical tests showed the families FASEO and FASEB are more efficient than FASE with respect to execution time in many experiments, also some methods achieving better accuracy results.


Comparative study of Fast Stacking Ensembles families algorithms

Incremental machine learning algorithms have been effective alternatives to deal with stream data.
The Hoeffding Tree framework is one of the most successful solutions for supervised online prediction tasks.
Although online regression tasks are present in several forms, and in many real-life problems, most of the research efforts have been devoted to classification.
Existing regression tree solutions have strong limitations, mainly regarding their memory usage and running time.
Hence, a new algorithm able to address these aspects in Hoeffding Tree Regressors is a relevant research issue.
We propose 2CS, a correlation-guided strategy to speed up Hoeffding Tree Regressor training.
2CS is conceptually simple and works by avoiding the exhaustive evaluation of all possible features as split candidates, as occurs in the existing solutions.
Moreover, 2CS can be easily merged into existing incremental tree solutions and online tree ensembles algorithms, such as bagging and boosting.
Throughout an extensive experimental evaluation, we show that the induction of 2CS-based models can be significantly faster than the traditional Hoeffding Tree Regressor algorithms, whereas retaining similar predictive power and memory use.

2CS: Correlation-guided Split Candidate\\Selection in Hoeffding Tree Regressors

Ontologies are used in Information Systems to manage and share knowledge pertaining to different domains, considering it is common that one particular piece of knowledge is modeled individually into different ontologies. Alignment processes are carried out towards interoperability; however, the results are not effective in most cases, requiring manual repair and review by domain experts. Thus, this work introduces a repair and review tool for the alignment of ontologies with the goal
of assisting experts in the alignment repair tasks. Using the technology acceptance model, a proof of concept was made to analyze the perception by domain experts regarding the tool’s usefulness, ease, and intention of use. From statistical analyses in terms of reliability, default path, inference, and correlation, it has been found that study participants had a positive perception of the tool’s usefulness and ease in the tasks of ontology alignment repair. Therefore, the tool and its functionalities
proved to be useful towards assisting domain experts in their tasks of reducing the semantic heterogeneity among ontologies pertaining to the same domain through alignment repair.

A Useful Tool to Support the Ontology Alignment Repair

Most ensemble methods for classification and regression tasks are based on a simple naive technique such as bagging, where all models' outputs are combined in a single last label. But in this paper, we evaluate a meta-learning approach that for each instance a different subset of classifiers is chosen.

Towards an Instance Level Meta-Learning Based Ensemble for Time Series Classification

A short-term electricity demand forecasting in the Southern Region of Brazil using the ARIMA Model and the Holt Exponential Smoothing Model (SEH)

I will present the Doclass, a free and open-source software for the Web that aims to assist in labeling and classifying large sets of documents. As a result, the first version of a system with an architecture composed of a mobile frontend that communicates with a backend through a REST API was published, with satisfactory performance evaluation by the applicant. Other results involve the use of active learning techniques to reduce human effort when performing the classification of documents, as well as the Uncertainty strategy to choose the document to be labeled.

Doclass: open-source software to support document labelingand classification

My presentation is about my solution to this year's KDD competition. There I cover the models, the methods of feature selection, validation, parameters optimization and also my final ensemble.

Solution KDD BR Competition 2020

O vídeo descreve o desafio da competição KDD-BR 2020, que trata da predição de períodos de indisponibilidade de aluguel de carros. Ele foi elaborado pela equipe de Ciência de Dados da empresa Localiza, que foi a proponente do problema.

KDD-BR: apresentação do problema pela empresa Localiza

KDD-BR 2020 solution using 1-dimensional convolutions.

Car availability prediction in rental angencies [KDD-BR 2020]

As commodities milho e soja são produtos consumidos em grande escala no mundo. As flutuações nos preços de mercado têm efeitos de longo alcance sobre os consumidores, agricultores e processadores de grãos. Assim, a previsão dos preços desses grãos tem atraído significativa atenção dos pesquisadores. Os modelos de previsão geralmente usam dados quantitativos de série temporal. No entanto, fatores qualitativos externos podem influenciar os dados em séries temporais, como eventos políticos, crises econômicas e o mercado de câmbio estrangeiro. Essas informações não são explícitas nos dados da série temporal e esses fatores podem influenciar a previsão dos valores das variáveis. Os dados textuais extraídos de notícias, fóruns e redes sociais podem ser uma fonte de conhecimento sobre fatores externos e potencialmente úteis para modelos de previsão de séries temporais. Alguns estudos apresentam técnicas de mineração de texto para combinar dados textuais com séries temporais. Porém, as representações existentes apresentam algumas limitações, como a maldição da dimensionalidade e atributos ineficazes. Este trabalho aplica métodos de pré-processamento em séries temporais e usa representações combinadas com dados textuais para prever o preço futuro do milho e da soja. Os resultados indicam que os métodos utilizados podem ser uma alternativa para melhorar o desempenho de previsão em tarefas de regressão.

Downloads

Next from BRACIS 2020

Comparative study of Fast Stacking Ensembles families algorithms

Similar lecture

Clustered Echo State Networks for Signal Observation and Frequency Filtering

.css-70qvj9{display:-webkit-box;display:-webkit-flex;display:-ms-flexbox;display:flex;-webkit-align-items:center;-webkit-box-align:center;-ms-flex-align:center;align-items:center;}Downloads

Next from BRACIS 2020

Comparative study of Fast Stacking Ensembles families algorithms

Similar lecture

Clustered Echo State Networks for Signal Observation and Frequency Filtering

Downloads