Human behavior, social interaction, and decision-making are all influenced by emotion, which is a basic, intrinsic state of mind. Owing to the Internet's explosive growth in the modern era, online social media (OSM) platforms are now widely used as a means of opinion and emotion expression. As a result of the development of artificial intelligence (AI) algorithms-driven natural language processing (NLP) methods, emotion detection (ED) from user-generated OSM data has become a heavily researched area. The brief, informal, and unstructured texts that are typical on micro blogging sites like Twitter make it difficult to extract useful features for spotting observable patterns. In this paper, we present a novel feature representation that can capture users' emotional states, taken from user-generated Twitter data. An enhanced method predicated on RF, The input representation is built using SVM, KNN, XGBOOST, and other techniques. It is made up of linguistic, sentiment, and stylistic elements that are taken from tweets. Using the new feature representation, a voting ensemble classifier with algorithmically optimized weights is presented to improve the accuracy of emotion detection. A benchmark Twitter emotion detection dataset is used to train and evaluate the suggested classifier. Each sample in the dataset is labelled with one of the six classes: fear, surprise, anger, sadness, joy, and love. The experimental results show that the suggested method achieves the highest accuracy and outperforms the most advanced classical machine learning-based emotion detection techniques.

RF, SVM, KNN, XGBOOST