Algorithms, data structures and visualization for sequence analysis and prediction

用于序列分析和预测的算法、数据结构和可视化

• 批准号: 435989-2013
• 负责人: FournierViger, Philippe
• 金额: $ 1.38万
• 依托单位: Université de Moncton
• 依托单位国家: 加拿大
• 项目类别: Discovery Grants Program - Individual
• 财政年份: 2015
• 资助国家: 加拿大
• 起止时间: 2015-01-01 至 2016-12-31
• 项目状态: 已结题
• 来源: https://www.nserc-crsng.gc.ca/ase-oro/Details-Detailles_eng.asp?id=574253
• 关键词: Algorithms; data; structures; visualization; sequence

Sequences are a common type of data that can be found in numerous domains such as biology, e-learning, market basket analysis, sensor networks, medicine, and meteorology. Such data is often stored in large sequence databases that need to be analyzed. However, analyzing it by hand is time-consuming. It is therefore crucial to develop algorithms for processing large amount of sequential data automatically. Two fundamental tasks that are performed on sequential data are (1) sequential analysis and (2) sequence prediction. The former consists of discovering patterns in sequences to gain insight on sequential relationships hidden in the data, to then use this insight for decision making. The latter consists of predicting the next elements of a sequence based on a set of previous sequences. These two problems are related and have wide applications in many domains. For example, sequence prediction can be applied to predict the next webpage that a user will visit based on previous users. But it could also be applied for text and music generation, and earthquake prediction. In this research program, we will develop new algorithms for these two tasks to improve the speed, memory usage and prediction accuracy. We will also develop visualization tools and algorithms that can provide advanced prediction capabilities such as performing online sequence predictions and quickly updating predictions. The results of this fundamental research will be applied in e-learning and other domains to be determined.

序列是一种常见的数据类型，可以在许多领域中找到，如生物学，电子学习，购物篮分析，传感器网络，医学和气象学。这些数据通常存储在需要分析的大型序列数据库中。然而，手工分析是耗时的。因此，开发自动处理大量连续数据的算法至关重要。 对序列数据执行的两个基本任务是（1）序列分析和（2）序列预测。前者包括发现序列中的模式，以获得隐藏在数据中的序列关系的洞察力，然后使用这种洞察力进行决策。后者包括基于先前序列的集合来预测序列的下一个元素。这两个问题相互关联，在许多领域有着广泛的应用。例如，可以应用序列预测来基于先前的用户预测用户将访问的下一个网页。但它也可以用于文本和音乐生成，以及地震预测。 在这项研究计划中，我们将为这两个任务开发新的算法，以提高速度，内存使用和预测精度。我们还将开发可视化工具和算法，以提供高级预测功能，例如执行在线序列预测和快速更新预测。本基础研究的结果将应用于电子学习和其他领域有待确定。