Optimizing Sentiment Classification: A hybrid approach combining NLP & advanced ML techniques

Abstract

  A cornerstone of numerous message mining applications is mood analysis. The writing has introduced an extensive variety of mood characterization procedures, including profound learning-based and conventional methodologies. Indeed, even while mood examination has been effectively applied in various business applications and huge achievements have been achieved utilizing the strategies currently being used, its precision might in any case be expanded. We tackle this theme by using a few Machine Learning (ML) and Deep Learning (DL)  techniques along with NLP approaches like count vectorizers and stopwords. Casting a ballot classifiers,  Long Short-Term Memory (LSTM), Recurrent neural networks (RNN), Gated recurrent units, random forests, Ada boost, stochastic gradient descent, k nearest neighbor, decision trees, multinomial naïve bayes, support vector machines, gradient boosting, multilayer perceptrons, and voting classifiers are among the calculations that are utilized. Of these, the democratic classifier has achieved the most elevated precision. also, utilizing dictionary to break down feeling utilizing extremity scores.