Sentiment analysis for movie recommendations: Harnessing opinion mining systems to analyze user reviews | ||||
| The Egyptian Statistical Journal | ||||
| Article 1, Volume 68, Issue 1, June 2024, Page 1-14 PDF (618.88 K) | ||||
| Document Type: Original Article | ||||
| DOI: 10.21608/esju.2024.252649.1024 | ||||
 View on SCiNiTO | ||||
| Authors | ||||
Doha Nor El-Deen  1; Rania Salah El-Sayed2; Ali Hussein3; Mervat Zaki4
| ||||
| 1Misr University for Science and Technology, Giza, Egypt | ||||
| 2the Department of Mathematics, Faculty of Science (girls branch), Al-Azhar University, Cairo, Egypt | ||||
| 3the Department of Mathematics at the Centre of Basic Science, Misr University for Science and Technology, Giza, Egypt | ||||
| 4Department of Mathematics, Faculty of Science (girls branch), Al-Azhar University, Cairo, Egypt | ||||
| Abstract | ||||
| Opinion mining systems now require sentiment analysis because of the massive amounts of data and opinions that are generated, shared, and sent every day through the Internet and other media. The major topic of this study is sentiment analysis for movie recommendations. There are too many reviews and comments to process manually. As a result, to process it successfully, we used user reviews of films (whether they were positive or negative) to create an overall assessment of reviews. A strategy must be developed to extract knowledge from the existing reviews and apply it more effectively. In this research work, two machine learning approaches are adopted, applied, and tested for the analysis and classification of user reviews. The first approach involves some supervised machine learning algorithms, namely Support Vector Machine (SVM) and Naïve Bayes (NB), which are applied based on feature selection algorithms, namely Term Frequency–Inverse Document Frequency (TF-IDF). The second approach is concerned with presenting a proposed model based on deep learning algorithms such as Artificial Neural Networks (ANN), Convolutional Neural Networks (CNN), and Long Short-Term Memory (LSTM) that are applied based on Word embedding techniques such as Glove that enable deep learning models to capture semantic relationships and contextual information. This enhances the models' ability to understand and analyze textual data. The test results demonstrated that LSTM outperforms other approaches despite CNN reporting accuracy better than ANN. Our models outperformed Support Vector Machine and Naive Bayes, as well. | ||||
| Keywords | ||||
| Artificial Neural Network; Long Short-Term Memory; Convolutional Neural Networks; Sentiment Analysis | ||||
|
Statistics Article View: 257 PDF Download: 333 |
||||
