A satin optimized dynamic learning model (SODLM) for sentiment analysis using opinion mining

• Published in: Social network analysis and mining 2023-08, Vol.13 (1), p.110
• Main Authors: Shanthi, D, Prabha, S. Santhana, Indumathi, N, Naganandhini, S, Shenbagavalli, S. T, Jayanthi, M
• Format: Article
• Language: English
• Subjects: Accuracy; Activities of daily living; Algorithms; Automation; Classification; Complexity; Data mining; Datasets; Deep learning; Digital media; Health services; Literature reviews; Machine learning; Mining; Natural language processing; Optimization; Performance assessment; Semantics; Sentiment analysis; Social media; Social networks; Vector space
• ISSN: 1869-5450; 1869-5469
• DOI: 10.1007/s13278-023-01114-8

Social media have grown more prevalent in daily life as a result of the quick development of internet technology. People use social media as a forum to share their thoughts, suggestions, and ideas. Sentiment analysis is used to grasp the emotional context of the language and to identify if the sentiment is good, negative, neutral, or any other emotional opinion. In the previous studies, there are several machine learning and hybrid learning techniques are developed for sentiment analysis and opinion mining. But, the majority of techniques limit with the main problems of high system complexity, low efficiency, increased training and testing time, and high false predictions. Therefore, the proposed work aims to develop a new framework, called as, Satin Optimized Dynamic Learning Model (SODLM) for sentiment analysis. In this system, the sophisticated algorithms such as Satin Bowerbird Optimization (SBBO) and Dynamic Ensemble Learning Classification (DyLC) models are used to accurately predict the sentiment from the given online review dataset. The stop words are removed, tokenized, stemmed, and lemmatized operations are carried out as part of the initial data preprocessing stage. The SBBO technique is then used to choose the relevant features from the normalized data in order to increase the classifier's ability to accurately predict sentiment. This technique also helps to reduce the complexity of training by reducing the dimensionality of features. Additionally, the prediction label is categorized using the DyLC technique into three categories: positive, negative, and neutral. In this study, the most well-known and cutting-edge datasets, including IMDB, medical services, Twitter airlines, and others, are used for validation. Several evaluation metrics are used in the performance assessment, and the results show that the SODLM offers the average accuracy. The performance assessment is carried out using several evaluation measures, and final outcomes indicate that the SODLM provides the average accuracy up to 99% for all datasets used in this research study.