Feature Selection with Hybrid Bio-inspired Approach for Classifying Multi-idiom Social Media Sentiment Analysis
Nenhuma Miniatura disponível
Data
2022-01-01
Autores
Silva, Luis Marcello Moraes [UNESP]
Valencio, Carlos Roberto [UNESP]
Zafalon, Geraldo Francisco Donega [UNESP]
Columbini, Angelo Cesar
Filipe, J.
Smialek, M.
Brodsky, A.
Hammoudi, S.
Título da Revista
ISSN da Revista
Título de Volume
Editor
Scitepress
Resumo
Social media sentiment analysis consists on extracting information from users' comments. It can assist the decision-making process of companies, aid public health and security and even identify intentions and opinions about candidates in elections. However, such data come from an environment with big data characteristics, which can make traditional and manual analysis impracticable because of the high dimensionality. The implications on the analysis are high computational cost and low quality of results. Up to date research focuses on how to analyse feelings of users with machine learning and inspired by nature methods. To analyse such data effectively, a feature selection through cuckoo search and genetic algorithm is proposed. Machine learning with lexical analysis has become an attractive alternative to overcome this challenge. This paper aims to present a hybrid bio-inspired approach to realize feature selection and improve sentiment classification quality. The scientific contribution is the improvement of a classification model considering pre-processing of the data with different languages and contexts. The results prove that the developed method enriches the predictive model. There is an improvement of around 13% in accuracy with a 45% average usage of attributes related to traditional analysis.
Descrição
Palavras-chave
Sentiment Analysis, Feature Selection, Cuckoo Search, Genetic Algorithm, Machine Learning, Social Media
Como citar
Iceis: Proceedings Of The 24th International Conference On Enterprise Information Systems - Vol 1. Setubal: Scitepress, p. 297-307, 2022.