Web2.0和普适计算技术的迅猛发展，加快了知识内容的生产和传播速度，拓宽了人们获取知识的途径，促进了学习范式从接受认知到建构认知再到分布式情境认知的转变，使得无处不在、无时不在、按需适应的在线学习成为可能。在线学习以其开放自由、不受时空限制的优势得到了学术界的广泛认同，正逐步成为一种常态化的学习方式。然而当前在线学习虽然拥有大规模的用户访问，但并没有大规模网络学习的发生，学习者的退出率高、参与度低。造成上述现象的原因既有来自学习者自身或外部环境的因素，又有来自课程和学习支撑平台的因素。而当前在线课程普遍采用“One Size Fits All”的资源组织模式、学习支撑平台无法提供动态适应学习者个体需求的学习支持服务才是导致学习者纷纷退出在线学习最根本、最直接的原因，也是当前在线学习领域最需要和最有可能破解的难题之一。因此，如何进行在线课程学习内容资源的个性化组织、如何为在线学习过程提供持续性适应的学习支持服务是解决上述难题的有效途径，而自适应在线学习过程的实现成为关键。 本研究以如何实现在线学习过程的适应性为核心关注点，以自适应在线学习认知地图模型构建为核心研究内容，重点解决在线学习过程中“One Size Fits All”的资源组织、无法提供动态适应学习者个体需求的学习支持服务的问题。针对上述问题，本研究主要在以下三方面开展了研究： 首先，设计了一种能够表征学习者知识结构和认知状态的适应性、开放性在线学习者模型——自适应在线学习认知地图模型。通过分析在线学习对适应性学习者模型所提出的三个方面需求，从适应的内容和适应的原理两个层面对适应性学习者模型进行思考，构建了自适应在线学习认知地图模型，阐述了自适应在线学习认知地图模型的组成要素及其之间的关系，介绍了自适应在线学习认知地图的生成框架及其涉及的五个关键环节。 其次，基于自适应在线学习认知地图模型，依托所在团队研发的学习元平台，综合利用J2EE技术、语义Web技术和可视化技术等设计并开发了自适应在线学习认知地图支撑系统。该系统包括知识点的垂直和水平层级结构、学习者的认知状态、学习者的知识结构、适应性的学习内容、适应性的学习活动、适应性的学习路径、适应性的学习同伴等可视化呈现功能模块，涉及学科知识图谱构建、学习交互数据采集、认知状态评估、关联关系挖掘、先修关系挖掘、知识掌握程度预测等核心技术。 最后，将自适应在线学习认知地图支撑系统应用于基础教育领域，并利用实践数据来检验和评估自适应在线学习认知地图的效果和影响。结果表明，自适应在线学习认知地图能够帮助学习者实时了解自己的知识结构和认知状态，发现自己的薄弱点和学习困难，快速获得与自己知识能力水平相匹配的学习资源、学习路径以及学习步调最一致的学习同伴，从而有效提升在线学习的效果和体验。 本研究在以下三方面进行了一定程度的创新： (1) 从知识结构化的视角，构建了一种能够表征学习者知识结构和认知状态的自适应在线学习认知地图模型，拓展了开放性学习者模型的内涵，突破了传统学习者模型封闭的局限。该模型具有结构化、可视化、复合性、差异性、动态性、关联性、生成性、过程性、开放性、持续适应性等核心特征，注重对学习者认知结构的过程变化进行表征，能够为在线学习的全过程提供反馈和个性化支持服务，推动了在线学习者模型研究的发展。 (2) 从持续性适应的视角，提出了从层级相似度和贝叶斯相似度两个维度来表征不同学习者之间学习认知地图相似度的计算方法，实现了个性化学习支持服务推荐的持续性适应，开辟了个性化推荐方法设计的新思路。 (3) 从工程化的视角，探索了基于学科知识点进行适应性学习的方法，并在小学信息技术课堂中进行实践，推动了个性化学习和针对性教学在基础教育领域中的应用，为智慧学习的开展、教育信息化的推进提供了一定的实践参考价值。 虽然本研究取得了一些成果，但仍存在一些不足，需要在后续研究中继续改进和完善，包括利用机器学习、文本挖掘等技术来实现学科知识点、学习内容与学习活动之间的自动关联，优化自适应在线学习认知地图支撑系统功能，扩大自适应在线学习认知地图支撑系统的应用领域和学科范围，利用脑电技术监测在线学习过程中的内部认知变化，探索造成实验班级认知负荷高于对照班级的原因。

The rapid development of Web2.0 and pervasive computing technology has accelerated the production and dissemination of knowledge, broadened the way of people obtaining knowledge, and promoted the learning paradigm changing from accepted cognition paradigm to constructive cognition paradigm and to distributed situated cognition paradigm. The change has made it possible for online learning to be ubiquitous and adaptive on demand. Online learning is widely recognized by the academic community for its openness and free from the constraints of space and time. Online learning is gradually becoming a normal way of learning. Although online learning has a large number of user visits, there is no real large-scale network learning. Online learners have a high dropout rate and low participation. The reasons for the above phenomena include factors from the learner's own or external environment, as well as from the curriculum and learning support platforms. However, the current adoption of the "One Size Fits All" resource organization model of online courses and the learning support platform failing to provide learning support services that dynamically adapt to individual learner needs are the most fundamental and direct reasons of dropout from online learning. Those problems are the most needed and the most likely to be solved in the current online learning field. Therefore, how to conduct the personalized organization of online course and how to provide continuous adaptive learning support services in the online learning process are the effective way to solve the above problems. The realization of the adaptive online learning has become the key. To solve the above problems caused by ‘one size fit all’ model and failing to provide learning support services that dynamically adapt to individual learner needs, this research focused on how to achieve the adaptability in online learning process, and took the adaptive online learning cognitive map model construction as the core research content. The following three aspects of research works were carried out: Firstly, this research designed an adaptive online learner model, named the adaptive online learning cognitive map, which could present the learner's knowledge structure and cognitive status. In view of the three demands of online learning for the adaptive leaner model, the research proposed the model by considering the content and principle to adapt, summarized the core characteristics of the model, and designed the elements and the relationship among them. In addition, the research introduced the generation framework of the adaptive online learning cognitive map and its five key steps. Secondly, the research designed and developed a support system for the adaptive online learning cognitive map based on the proposed model and the Learning Cell knowledge community developed by the team by comprehensive utilization of J2EE technology, semantic Web technology and visualization technology. The core functions of the support system included visualized presentation of the vertical and horizontal structure of knowledge, the learner's cognitive status, the learner's knowledge structure, the adaptive learning path, the adaptive learning content, the adaptive learning activity and the adaptive learning partner. Furthermore, the support system involved several core technologies, such as the construction of the discipline knowledge graph, the gather of learning interactive data, the assessment and prediction of cognitive status, the mining of association relationship and prerequisite relationship. Finally, the research applied the adaptive online learning cognitive map model in the field of basic education and the practice data to evaluate the application effect of the model. The practices result demonstrated that the adaptive online learning cognitive map could help learners understand their own knowledge structure and cognitive status in real time, and find their own weaknesses and learning difficulties. Meanwhile, the adaptive online learning cognitive map could present the adaptive learning resources, adaptive learning path and adaptive learning partner that match their knowledge level and learning pace, which effectively enhance the effectiveness of online learning and experience. Based on the above research work, this research has obtained a certain degree of innovation in three aspects as following: (1) From the perspective of knowledge structuration, this research constructed an adaptive online learning cognitive map model that could present the learners' knowledge structure and cognitive status, which expanded the connotation of the open learner model and broke through the limitation of the traditional learner model. The adaptive online learning cognitive map model had ten kernel characteristics, such as structuralization, visualization, complexity, differentiation, dynamism, relevance, generativity, processness, openness and continuous adaptability. Paying attention to the changes of learners’ cognitive structure during online learning, the proposed model could provide personalized feedback and support service for the whole learning process and promoted the development of the online learner model. (2) From the perspective of continuous adaptation, the research proposed a method to compute the similarity of learning cognitive maps between different learners in the aspects of hierarchical similarity and Bayesian similarity. The similarity could be utilized for the continuous adaption of personalized learning support service, which opened up a new approach to design the method of personalized recommendation. (3) From the perspective of engineering, the research explored the method of adaptive learning based on subject knowledge points, and applied it in the information technology course of primary schools. The research promoted the application of personalized learning and teaching in elementary education. This research provided a practical reference for the development of smart learning and the promotion of educational information. Although this research had obtained some preliminary achievements, there are still some limitations, which are worth improving in the follow-up study. It includes: a) Using machine learning, text mining to realize the automatic association between subject knowledge points and learning contents and learning activities. b) Optimization the support system. c) Expanding the application field and discipline of online cognitive map. d) Using EEG technology to monitor the internal cognition changes in online learning process. Moreover, e) Exploring the causes of the experimental class having higher cognitive loads than the control class.