Suggestion analysis involves computationally analysing textual expressions like tips, advices, recommendations etc in online reviews. This work utilizes a deep learning framework using models, such as CNN, LSTM, LSTM-CNN, and CNN BiLSTM-GRU architectures, with BERT representations. The evaluation of these models is conducted using Software Forum Domain related to electronics reviews and TripAdvisor domain associated with hotel reviews. To assess the model accuracy, precision, recall and F1score metrics are calculated. Notably, transfer learning hybrid models have demonstrated superior performance in suggestion analysis and particularly BERT-CNN BiLSTM-GRU model.