<p>Customer reviews on e-commerce platforms provide crucial insights into user experience and product quality. However, accurately predicting review ratings based on textual feedback has become an essential task in understanding customer satisfaction. This study aims to develop a novel deep learning-based model that can accurately predict numerical ratings from unstructured product review texts. In this context, we propose RateNet, a hybrid deep learning model combining 1D CNN and BiGRU layers for review rating prediction. The 1D CNN layers capture local semantic patterns from reviews, while the BiGRU layers learn long-term dependencies, aiming to enhance the rating prediction accuracy. The proposed method leverages three different feature extraction techniques, viz. TF-IDF, Word2Vec, and GloVe, to encode text reviews into structured representations. The effectiveness of the proposed model is tested on the Amazon 5-core software review dataset using metrics such as accuracy, precision, recall, and F1 score. The experimental findings reveal that the RateNet model, when paired with TF-IDF features, attains an accuracy of 86.60%, substantially exceeding the performance of existing methods, including traditional machine learning models, deep learning architectures, and transformer-based models. Additionally, recall, precision, and F1 score values consistently surpass advanced existing methods. Confidence interval tests provide further validation of the proposed RateNet model’s performance, emphasizing its statistical significance, yielding the lowest error rate, and the smallest margin of error with TF-IDF features. The proposed RateNet model demonstrates robust performance in predicting review ratings from text, particularly when coupled with TF-IDF features. Its hybrid architecture offers improved performance and generalization over existing methods.</p>