在线学习是正在快速发展的一个领域。推荐系统能为用户提供个性化的推荐服务，目前基于深度神经网络(DNN)的模型性能最好。然而，对在线学习领域的推荐系统的研究数量少，且研究质量不高、前沿性不足。因此，本项目拟对此开展深入研究，包括自适应森林(SAF)研究、基于SAF的推荐模型研究、特征工程研究、超参数的调优算法研究等。本项目的创新之处有：1)不同于现有的面向在线学习的推荐系统的研究思路，本项目拟建立一种非DNN框架下的机器学习模型——SAF并基于此构建推荐模型，预期能达到与基于DNN的推荐模型相同或更好的性能并克服其一些不足，这在研究思路上具有较高的创新性；2)创新性地开展面向在线学习的推荐系统的特征工程研究，弥补了目前尚无这一研究的不足；3)引入“基于种群的训练”思想，建立一种新的超参数调优算法。本项目的研究涵盖机器学习任务(即：推荐)的三个重要方面，预期能取得较好的推荐性能。

E-learning is one of the rapid developing fields in education. Recommender systems can provide the users with personalized recommendation services, for which Deep Neural Network (DNN)-based models achieve the best performance at present. Nevertheless, research on recommender system for E-learning not only is in tiny quantity, but also is in low quality and is not at the cutting-edge. Therefore, this project is to conduct in-depth research on it, including research on Self-Adaptive Forest (SAF), research on SAF-based recommendation model (RM), research on feature engineering, and research on hyperparameter tuning algorithm. The innovation of this project lies in: (1) Different from current research route of recommender system for E-learning, this project intends to develop a non-DNN model, i.e., SAF, and construct an RM based on it, which is expected to achieve the same or better performance compared with DNN-based RMs and at the same time overcome some shortcomings. The above-mentioned research route has relatively high novelty. (2) This project will innovatively conduct research on feature engineering of recommender system for E-learning, which overcomes the insufficiency that there lacks research on this topic. (3) By introducing the idea of population-based training, this project will develop a new hyperparameter tuning algorithm. Research of this project covers the three important aspects of a machine learning task (i.e., recommendation), thus the RM to be constructed is expected to have good performance.