Small: The MCDB Database System for Managing and Modeling Uncertainty

管理和建模不确定性的MCDB数据库系统分析人员经常使用统计模型来“猜测”与数据库中存储的数据相关的未知，不准确或缺失的信息。例如，一家制造商的分析师可能想知道，“如果去年我的利润率提高了5%，我的利润会是多少？”这个问题的答案取决于更高的价格对每个客户需求的影响程度，这无疑是通过应用一些统计模型来猜测的。MCDB项目关注的是一个原型数据库系统的设计和实现，该系统被称为“蒙特卡洛数据库系统”，简称“MCDB”。MCDB允许专家级分析师或统计学家将任意随机模型附加到数据库数据上，以便“猜测”未知或不准确数据的值，例如每个客户看不见的需求函数。这些随机模型驻留在数据库中，并且始终是最新的，因为它们是根据数据库的当前状态参数化的（在上面的例子中使用每个客户最近的购买）。该项目解决了许多关键的智力和科学挑战。其中大多数都与以下事实有关：出于性能原因，不可能实现1tb数据仓库的1000个随机实例，并按顺序查询它们中的每个实例。人们正在考虑避免这种物质化的新方法，例如跳过产生永远不会用于回答特定查询的数据的蒙特卡洛试验。该项目还考虑了统计方面的挑战，例如生成远在答案分布尾部的数据库实例，这对于风险评估等特定应用程序是必要的。欲了解更多信息，请访问http://mcdb.cs.rice.edu。

The MCDB Database System for Managing and Modeling UncertaintyAnalysts working with large data sets often use statistical models to``guess'' at unknown, inaccurate, or missing information associatedwith the data stored in a database. For example, an analyst for amanufacturer may wish to know, "What would my profits have been ifI'd increased my margins by 5% last year?" The answer to thisquestion depends upon the extent to which the higher prices would haveaffected each customer's demand, which is undoubtedly guessed via theapplication of some statistical model.The MCDB project is concerned with the design and implementation of aprototype database system called the "Monte Carlo Database System," or"MCDB" for short. MCDB allows an expert-level analyst or statisticianto attach arbitrary stochastic models to the database data in order to"guess" the values for unknown or inaccurate data, such as eachcustomer's unseen demand function. These stochastic models reside inthe database, and are always up-to-date in the sense that they areparameterized on the current state of the database (using eachcustomer's most recent purchases in the above example).The project attacks a number of key intellectual and scientificchallenges. Most of these are related to the fact that forperformance reasons, it is not possible to materialize one thousandstochastic instances of a one terabyte data warehouse, and query eachof them in sequence. Novel methods for avoiding such materializationsare being considered, such as skipping Monte Carlo trials that producedata which will never be used to answer a specific query. The projectalso considers statistical challenges, such as generating databaseinstances that fall far out in the tail of the answer distribution,which is necessary for specific applications such as risk assessment.Further information is available at http://mcdb.cs.rice.edu.