E-commerce Product Review Analysis based on Multi-class Support Vector Machine

Recently, an extensive number of consumers can choose their best options from automated retailers and compare products in online stores. Sentiment analysis is frequently employed in applications meant to demonstrate and provide client service as the voice of the user. On E-commerce websites, user reviews provide helpful information about the product. Sentiment analysis of the text reviews helps forecast product sales by assessing user sentiment regarding the product. The Word Embedding Attention approach assigns additional weight to words that have a strong association with a specific class. Sentiment analysis machine learning models provide a thorough depiction of capabilities and far surpass conventional feature-based methods. The obj ective of this research is to improve sentiment analysis performance by creating a weighted ensemble with a Multi-class Support Vector Machine (MSVM) model utilising novel word embedding techniques. Because of the weighted ensemble's higher generalisation skills, MSVM produces superior results. When comparing existing methods for sentiment analysis, the Weighted ensemble with MSVM has produced higher results in terms of accuracy, precision, f-score, and recall. The Weighted ensemble with an MSVM has achieved 99.87 % accuracy, 99.39% precision, 99.03% f-score, and 99.690/0 recall in sentiment analysis when comparing existing methods such as LSTM+FL, APSO-LSTM, ABCDM, SSentiA, and Hybrid CNN+LSTM.