A Hybrid Deep Learning Framework for Improving User Review Classification for Usability and Security& Privacy
Kiranbir Kaur, Kuljit Kaur Chahal · IEEE Access · 2025
The wide spread of mobile applications forced people to perform their daily activities online. Mobile application developers heavily depend on user feedback in the form of reviews, as it helps them identify problems with their apps and ensures they currently meet user expectations. User reviews are inherently informal, disorganized, and unstructured. It is challenging to extract and classify the information on relevant issues due to the vast number of reviews. To expedite the classification of reviews, many researchers implemented machine-learning approaches. This work proposes a deep learning-based framework, MAUSPC, which utilizes RNN-LSTM to classify user reviews of mobile learning applications (apps), assessing usability, security& privacy. This approach leverages deep learning’s ability to extract useful information and automatically create classifications. The proposed framework builds on the latest developments of the GPT-3 Transformer by utilizing a text embedding approach. This method offers a high-quality depiction that can enhance the efficacy of detection. Furthermore, we employed the hybrid approach, which combined RNN-LSTM models into a single model that performed noticeably better than the sum of its components. The dataset of mobile learning applications was divided into two categories—usability and security & privacy—was used to conduct the model’s experimental assessment. With an accuracy of 98.59%, the results demonstrated performance that surpassed all other approaches.