Exploring ML Methods for Sentiment Analysis in Text Data

Data Analysis has become an important part of our life, as it helps us to draw useful information, decision-making conclusions on any particular raw data. One component of digital data analysis is sentiment analysis. Sentiment analysis evaluates the emotional tone of textual data and classifies it as neutral, negative, or positive. In business, this analysis is frequently utilized for market social media monitoring, market research, consumer feedback and many more. In order to automate the procedure, machine learning models that recognize patterns in labeled data are frequently used. The objective is to derive meaningful insights from textual data so that decisions can be made with knowledge, effectively. In order to provide important insights into public opinion and behavior, this study uses machine learning techniques to investigate the sentiments conveyed in Twitter data. A wide range of Twitter posts, including ones about trending hashtags, events, and subjects, are gathered as part of the process. Sentiment analysis is done using machine learning algorithms. Sentiment analysis and examples of human language are used to train machine learning models. Following this training, machine learning is used by sentiment analysis software to evaluate and rank human language according to its previous training. The current study proposes a technique that best analyzes the sentiments of the text used in social media is gradient boosting classifier. With this method, an additive model can be built in a step-by-step, forward fashion and any differentiable loss function can be optimized. There are n-class regression trees that are fitted on the negative gradient of the loss function, which could be the multiclass log loss or the binary loss. This process is repeated in each step. In the particular case of binary classification, just one regression tree is induced. The computational results demonstrates that gradient boosting approach with an accuracy of 93.62% which outperforms is preferable to other classifiers.