Mining Interesting Useful Patterns

频繁模式挖掘是一项重要的数据挖掘任务，它可以找到频繁同时出现的项目集。许多现有算法从精确数据（例如超市交易）中找到频繁模式，其中数据集的内容是精确已知的。然而，在现实生活中，由于遗传测量误差或采样频率等因素，数据不精确或不确定（例如传感器数据、医学测试结果）。尽管存在不确定性，但这些数据包含丰富的有用知识。在过去的几年里，我开发了一些算法，使用概率方法从不确定的数据中查找频繁模式，其中每个事务中的项目都被假定为独立的。然而，这个假设在许多现实生活中可能并不成立。因此，我提出了一个研究计划，其目标是构建一个探索性、高效、用户友好且强大的挖掘框架，该框架由从数据流和/或不确定数据中挖掘用户感兴趣的有用模式的系统组成。具体来说，我计划 (i) 探索非概率方法来寻找频繁模式，(ii) 放宽上述假设，以便处理更现实的情况，即每个不确定交易中的项目可能相关，(iii) 在挖掘过程中纳入用户偏好，以便让用户找到用户感兴趣的其他有用（频繁或不频繁）模式，(iv) 进一步提高性能，以便为用户提供实时响应， (v) 开发可视化分析工具，使用户能够可视化和分析精确（或不确定）数据的静态（或动态）数据集。因此，对于这个拟议的研究计划，我和我的总部将开发新的数据挖掘技术来挖掘有趣的有用模式。这反过来又提高了该领域研究人员的知识。此外，我还计划将所提出的系统应用于各种现实生活中的应用（例如，挖掘网络数据、电信数据、农业气象数据和来自社交网络的推文），以证明所提出的系统在挖掘有趣的有用模式时满足应用程序用户的科学/业务需求的有效性。

Frequent pattern mining is an important data mining task that finds sets of frequently co-occurring items. Many existing algorithms find frequent patterns from precise data (e.g., supermarket transactions), in which the contents of datasets are precisely known. However, there are real-life situations in which data are imprecise or uncertain (e.g., sensor data, medical test results) due to factors like inherited measurement inaccuracies or sampling frequency. Despite their uncertainty, these data contain a rich set of useful knowledge. Over the past few years, I have developed algorithms that use probabilistic approaches to find frequent patterns from uncertain data, in which items in each transaction are assumed to be independent. However, this assumption may not hold in many real-life situations. Hence, I propose a research program with an objective to build an exploratory, efficient, user-friendly, and powerful mining framework--which consists of systems that mine useful patterns that are interesting to users from data streams and/or uncertain data. Specifically, I plan to (i) explore non-probabilistic approaches in finding frequent patterns, (ii) relax the above assumption so as to handle more realistic situations where items in each uncertain transaction may be related, (iii) incorporate user preferences in the mining process so as to allow users to find other useful (frequent or infrequent) patterns that are interesting to users, (iv) further improve performance so as to provide users with real-time responses, (v) develop visual analytics tools so as to enable users to visualize and analyze static (or dynamic) datasets of precise (or uncertain) data. Consequently, for this proposed research program, I and my HQP would develop new data mining technology for mining interesting useful patterns. This, in turn, advances knowledge of researchers in the field. Moreover, I also plan to apply the proposed system to various real-life applications (e.g., mining Web data, telecommunication data, agro-meteorological data, and tweets from social networks) so as to demonstrate the effectiveness of the proposed systems in addressing scientific/business needs of the application users when mining interesting useful patterns.