With the rapid development of big data technology and machine learning algorithms, intelligent autonomous learning systems show great potential for application in the field of education. This study aims to construct an intelligent autonomous learning system based on machine learning algorithms to achieve personalised learning and improve the quality of education. The system adopts a layered architecture, including perception layer, processing layer, cognitive layer and action layer, covering the whole process from data collection, pre-processing, feature extraction to decision making and action execution. By integrating multiple machine learning models, the system is able to analyse students’ learning behaviours, affective states and academic performance, and provide customised learning resources and teaching strategies. Experimental results show that the system has significant advantages in improving students’ learning efficiency, enhancing learning motivation and optimising the allocation of educational resources. The study also explores the challenges that the system may encounter in actual deployment and proposes solutions accordingly. The innovation of this thesis is that it proposes an educational system architecture that combines multimodal data and deep learning techniques, which provides a new direction for the future development of educational technology.