Application of Support Vector Machine and Convolutional Neural Network for Sentence-Level Sentiment Analysis of Companies Products Review

Akande, Oluwatobi Noah and Ayoola, Joyce and Misra, Sanjay and Ahuja, Ravin and Agrawal, Akshat and Jonathan, Oluranti (2022) Application of Support Vector Machine and Convolutional Neural Network for Sentence-Level Sentiment Analysis of Companies Products Review. In: Advances in Electrical and Computer Technologies(ICAECT 2021), 26 June 2022, Online.

PDF Download (137kB)

Abstract

The longevity of any company’s sustainability is largely dependent on its’ capability to satisfy customers’ needs. Hence, many industrial analysts and researchers have acknowledged the impact of customers’ reviews, which is the most common approach and tool used to ascertain the level of satisfaction or dissatisfaction of. To determine the sentiment polarity and classification of product reviews, several authors have employed different tools with varying degrees of accuracy when compared to other models. In this study, a hybridized CNN-SVM algorithm model was developed for sentence-level sentiment analysis on musical products reviews on Amazon. A total of 44,463 training samples and 19,056 samples were used for testing and validating the model. Based on the performance metric of the hybrid models, CNN had an accuracy of 85.38%, a precision of 90.56%, a recall of 95.14%, and an AUC of 0.836, whereas, SVM had an accuracy of 85.74%, a precision of 85.62%, a recall of 100%, and an AUC of 0.5. It was observed that there are more positive sentiments and satisfaction obtained from these products as against the negative sentiments. Hence, the level of satisfaction on these products reflected on the analyzed sentiments.

Item Type:	Conference or Workshop Item (Paper)
Subjects:	Q Science > QA Mathematics > QA75 Electronic computers. Computer science
Divisions:	Faculty of Engineering, Science and Mathematics > School of Electronics and Computer Science
Depositing User:	Patricia Nwokealisi
Date Deposited:	19 Jul 2024 15:15
Last Modified:	19 Jul 2024 15:15
URI:	http://eprints.covenantuniversity.edu.ng/id/eprint/18271

Actions (login required)

View Item