<p>Social media is an influential platform to communicate with the public to share their emotions or sentiments like thoughts and observations regarding every matter or article that contains a huge amount of unstructured information. Sentiment analysis makes use of text mining and natural language processing to detect and extract subjective information from written content. To increase the profit in business insight, the organization need to investigate and study the sentiments of people. However, many of the words have more than one meaning so it becomes more challenging to analyze sentiment in this situation. Traditional sentiment analysis methods often struggle with the fundamental ambiguity and complexity of human language. To address these challenges, this paper introduces a method namely the Fuzzy Hierarchical Convolutional Neural Network (Fuzzy HCN-Net). Initially, the review document is used as an input that is given into the tokenization phase, which is performed using Bidirectional Encoder Representations from Transformers (BERT). After that, the aspect term extraction (ATE) is utilized to extract the relevant terms from the text. Further, feature extraction is performed to extract various features that include punctuation marks, all caps, hashtags, question marks, emoticons, Term Frequency-Inverse Document Frequency (TF-IDF), sentence length, and word2vec. Finally, sentiment classification (SC) is performed by the proposed hybrid deep learning (DL) named Fuzzy HCN-Net. The metrics, such as accuracy, F1-score, and precision are considered to validate the performance of the Fuzzy HCN-Net model and attained maximum values of 91.5, 90.4 and 90.6%, respectively.</p>