Uncertainty Handling in a Tabular Representation

Abstract

Abstract-Nowadays a lot of uncertainty appears due to the burst of data on the internet. This challenges the traditional data processing methods because of its volume and variety. Such uncertainty is hard to be represented in simply continuous distribution functions when it is too hard or complicated to be obtained. Considering the realistic situations that people may be interested in queries falling into a scope rather than a specific value, we present a new data model, called N-DB model, where the attribute value is represented with a tabular form in an increasing order, denoted N-table (i.e., a set of ordered pairs). This model can deal with queries in a scope efficiently, such as "movies with reputation level bigger than 5". We modify the relational algebra, including the aggregation query, and show the query processing in a running example. We also define an uncertain measurement (-precision) to measure the precision and information of the N-table, which can be used in the computation of precision requirements. Through experimental results, we find queries can be evaluated efficiently by searching in the N-tables, and are returned with confidence intervals.