This research employs a convolutional neural network (CNN) with long short-term memory (LSTM) to analyse and predict the behaviour of users of mobile devices, utilising a dataset comprising 700 users. The model combines the strengths of convolutional neural networks (CNNs) in spatial feature extraction and long short-term memory (LSTM) networks in temporal sequential analysis. The results demonstrate that the model exhibits excellent performance, with 92% accuracy, 89% precision, 91% recall, and 90% F1 score. The temporal pattern analysis revealed significant variation between the user classes, with the intensive class showing consistently high usage, averaging 300 minutes per day. The key factors influencing the user experience were identified as app usage time (25%), screen on time (22%), and battery consumption (18%). The segmentation of users resulted in the identification of five distinct groups, with Segment 2 exhibiting the highest usage level (6.2 hours per day) and Segment 5 displaying the lowest (1.3 hours per day). The strong correlation (0.89) between app usage time and screen time serves to confirm the importance of optimising the performance of apps. These findings provide a basis for more effective service personalisation and more targeted app development, thereby paving the way for the optimisation of the user experience on mobile devices.