Custom web interfaces based on decision trees: significant improvement in usability with lowcomputational cost

Custom web interfaces based on user characteristics is a key challenge to improve user experience without significantly increasing computational complexity. This study proposes a solution based on decision trees trained with objective and subjective interaction indicators, in order to automatically assign user profiles and customize the corresponding interface. The proposal was validated in an experimental environment, focused on a functional module of a web system, where a substantial improvement in usability was observed: the average score on the SUS scale increased from 56.20 to 76.68, surpassing the satisfaction threshold (SUS > 68) and confirming the initial hypothesis. Likewise, a high acceptance of the assigned user profile was recorded (90% in the sample and 83% in the population), and a low use of the profile change functionality was observed (5--10%), suggesting conformity with the initial personalization. The solution was shown to be efficient and sustainable, standing out due to the automation of the data collection process for model training, the effective classification of users, and the low associated computational cost. In addition, a profile management module and a guided tour were implemented, which strengthen user autonomy and reduce the need for technical support, thus increasing the sustainability of the proposed software solution.