مدل‌سازي رفتار كليك كاربر با استفاده از الگوريتم‌هاي يادگيري ماشين براي بهبود تجربه‌كاربر

Modeling user click sequences in a mobile application can significantly enhance researchersʹ understanding of user experience by analyzing behavior across different pages of the app. By predicting which element a user is likely to click on next, it is possible to continuously improve the interface and overall user experience. This research has practical implications for increasing productivity and quality in user experience testing, leading to better software experiences and ultimately enhancing product user-friendliness and user satisfaction. In this research, after reviewing previous research and eva‎luating its strengths and weaknesses, over 35,000 click events from 50 users interacting with a mobile application in the context of employee recognition software were collected and preprocessed. The data include user IDs, clicked element names, page names, and click time, all of which were used as inputs to the model. The study focused on predicting the userʹs next click using LSTM and Transformer algorithms to understand the relationships within user click sequences. In this research, an accuracy of 88% was achieved with the LSTM model and 86% with the Transformer model using the "Top-3 Accuracy" metric. Additionally, an accuracy of 74% with the LSTM model and 67% with the Transformer model was achieved using the "Top-1 Accuracy" metric.

5