Due to the exponential growth of textual information on the internet, online monitoring and mining of textual data have become a prominent task for researchers, which requires a deeper understanding of text classification algorithms. Several machines and deep learning algorithms have performed well in natural language processing for textual classification. These classification algorithms extract helpful information from textual resources and automatically classify them into multiple predefined categories based on their content and subject matter. In this research paper, a comparative review sentiments analysis of flip kart products has been done by using different deep learning algorithms like Gated Recurrent Unit (GRU), Long Short-Term Memory (LSTM), Bidirectional Encoder Representations from Transformers (BERT) and Robustly Optimized BERT Approach (RoBERTa) on a given dataset in which their efficiency is analysed and compared. and from obtained experimental RoBERTa model scored highest prediction accuracy, with 86.94%.