Multi stage sentiment analysis for product reviews on Twitter using optimized machine learning algorithm

Modern consumer feedback is located on societal networks, and therefore the activity of such sites as Twitter is important for product reviews. This work explores the feasibility of using machine learning algorithms to classify sentiments in the context of tweets about products, providing marketing departments with richer insights into customer opinions. The proposed approach is that of an optimal multi-stage framework of sentiment classification that combines both conventional machine learning models, i.e. Support Vector Machines (SVM), Naive Bayes, and the Random Forest, and deep learning, i.e. Long Short-Term Memory (LSTM) networks. The novelty of the present work is the optimization and comparison of the application of these methods and the determination of the most effective sentiment classification method of the product-related discussions. A set of 5200 English tweets gathered within one months was taken into consideration, including positive, negative and neutral comments on opinions. The results show that machine learning can be useful in extracting and analyzing unstructured textual data from social media, supporting the use of social media sentiment analysis. Furthermore, the research addresses the possibility of applying sentiment analysis as an effective business strategy for enhancing client approaches and marketing appeals. By analyzing consumer attitudes towards products and services indicated on the site, companies can improve products and services, resolve issues, and gain consumer loyalty. This study provides valuable insights into the relationship between customer opinions and business strategies in today's competitive environment.