Off-line and integrated data mining for intelligent systems

万维网的广泛和易用性帮助使基于Web的应用程序成为软件行业的常规做法。当涉及到电子商务和远程教育等分布式应用程序时，这种做法更加明显。对于这些应用程序，必须更好地了解用户的在线行为，以便改进应用程序(即在电子商务的情况下增加利润，在电子学习的情况下改善学习体验)。基于Web的应用程序在Web访问日志中跟踪在线资源访问，该日志通常变得非常大。从基于网络的活动，特别是电子学习产生的数据的不同之处在于绝对的复杂性和巨大的数据量，以及不可能简单地提取信息的事实。有关用户行为的相关信息必须通过交互式数据挖掘和相关的数据可视化技术来推断。这项研究的重点是设计特定于基于网络的信息系统的背景下的数据挖掘技术，特别是在电子学习中的应用。这项研究将解决(1)离线数据挖掘(即在收集数据后执行的数据挖掘)和(2)集成数据挖掘(即对数据流进行实时数据挖掘)的问题。离线数据挖掘的目标是设计新的算法(如在聚类、分类、关联分析、序列分析等方面)。以及新的可视化技术，用于可视化网站路径遍历以及资源访问和使用，以帮助网络设计者和教育工作者在电子学习的情况下分析评估在线活动，并最终评估用户(即学习者)。集成数据挖掘的目标是设计智能软件代理，监控在线学习者的活动，并相应地动态调整和重组在线资源，并根据社区中类似学习者以前的成功学习行为自动向学习者推荐学习活动。

The widespread and ease of use of the World-Wide Web have helped make web-based applications a regular practice in software industry. This practice is even more pronounced when it comes to distributed applications such as electronic commerce and distance learning. With these applications, there is an important need to better understand the user on-line behaviour in order to improve the application (i.e. increase profit in the case of e-commerce, and improve the learning experience in the case of e-learning). Web-based applications track on-line resource accesses in a web access log, which typically becomes very large in size. What distinguish the data that result from web-based activities in general and e-learning in particular are the sheer complexity and the vast size of data, and the fact that simple information extraction is not possible. Relevant information about user behaviour must be deduced by interactive data mining and associated data visualization techniques. This research focuses on devising data mining techniques specific to the context of web-based information systems and in particular with application in e-learning. The research will address issues for (1) off-line data mining (i.e. data mining performed after the collection of data), and (2) integrated data mining (i.e. data mining performed real-time on data streams). The objectives for off-line data mining are to devise new algorithms (such as in clustering, classification, association analysis, sequence analysis, etc.) and new visualization techniques for visualizing website path traversals and resources access and usage to help web designers and educators in the case of e-learning assess analytically the on-line activities and eventually evaluate the users (i.e. learners). The objectives for integrated data mining are to design intelligent software agents that monitor on-line learners activities and dynamically adjust and restructure the online resources accordingly, as well as automatically recommend learning activities to learners based on previous successful learning behaviour of similar learners in the community.