How to SENTIMENT ANALYSIS ON SOCIAL MEDIA

Social websites, electronic media, and blogging services are all growing in popularity, and they generate a large number of user communications such as customer comments, reviews, and opinions. These messages are used to assess the quality of a variety of items and services, but they are quite many. As a result, tracking and recognising customer attitudes is tough. As a result, Sentiment Analysis (SA) or Opinion Mining (OM) was developed, which entails gathering data from the web and analysing opinions using Natural Language Processing (NLP) techniques.It extracts subjective information, analyses opinions, and measures positive and negative reviews before making a judgement regarding a specific entity (i.e. organisation, public, product).

The research's major goal is to create an automated system that can analyse a big dataset of movie evaluations for SA or OM based on aspect. NLP gathers opinions, which are then categorised as favourable, neutral, or negative. The study optimises marketing promotions and assists customers in selecting the best product for their needs. Various machine learning and swarm intelligence optimization strategies are used in this study to evaluate sentiment analysis for covid-19 data. New sorts of sentiment analysis models were investigated in this study.

Initially, the topic of extraction from Twitter and Facebook was explored, for which an algorithm was presented for recognising and extracting tweets related to the policy passed as a keyword. In terms of smart comment extraction and social media, the process was assessed through a comparison of alternative techniques to data extraction. Then, by introducing the notion of negation handling mechanism, the basic classification algorithms were improved.The model given in this thesis was developed using probabilities, with the data first being trained and then being tested in relation to the policies implemented by the Indian government. These requirements have sparked a new line of inquiry into the summarization of viewpoints. To construct an abstractive summary, an algorithm has been devised that combines the notions of extractive and abstractive versions. Throughout the inquiry, comparisons to existing methodologies have been made.

For efficient sentiment analysis, a hybrid support vector machine (SVM) with chicken swarm optimization (CSO) algorithm has been developed in this study. The possible features are extracted using Part-Of-Speech (POS) tagged text in this technique. The efficient characteristics have been chosen based on the opinion score, and the opinion scores and ranks for each noun are provided by the Chicken Swarm Optimization (CSO) scheme. The highest opinions of characteristics are then forecasted using the maximum opinion score scheme. The anticipated attributes are passed on to the classification step, where the Interactive Dichotomizer version 3 (ID3) is used to identify moods such as positive, neutral, and negative.

MATLAB was used to analyse these labour performances using movie review data. When compared to existing sentiment analysis schemes such as SVM, Logistic Regression, and Neural Network, simulation results show that the proposed HSVMCSO scheme achieves a high accuracy rate of 92.80 percent, sensitivity rate of 93.20 percent, specificity rate of 90.42 percent, precision rate of 95.38 percent, recall rate of 90.24 percent, and F-Measure rate of 89.78 percent. When compared to existing schemes, the suggested HSVMCSO achieved a processing time of 29.12s and a lower processing cost.

Conclusion

The entire research project was focused on developing a COVID-19 sentiment analysis technique based on machine learning and swarm intelligence, as well as effective feature extraction approaches. The major goal of this project is to reduce errors in deriving opinions from people's emotions during the COVID-19 epidemic. To improve the efficiency of opinion classification by increasing precision and reducing processing time.