Accessing imprecise data : an approach based on intervals

In many real world applications (even in banking), imprecise data is a matter of fact. However, classic database management systems provide little if any help in the management of imprecise data. We are applying methods from interval arithmetic, epsilon serializability, and other related areas to help the application designers in the management of naturally imprecise data. Our approach includes operators on imprecise data that give bounds on the result imprecision and algorithms that constrain the imprecision propagation in the database. 1 Introduction Traditional database management systems provide support only for precise data, though in the physical world data is often imprecise. An important class of examples is the scientific data such as incomplete recording of data, instrument noise, measurement error, computational model imprecision, and data aggregation of one kind or another. Another example is the fuzzy data managed by Epsilon Serializability algorithms PL91, DP93. In t...