Actioning Big Data

Abstract We propose to construct probability distributions for data subsets selected randomly from big data. A sequential decision support system will generate local decision support for every random data subset. At the end... [ view full abstract ]

Abstract

We propose to construct probability distributions for data subsets selected randomly from big data. A sequential decision support system will generate local decision support for every random data subset. At the end of the random data extraction process we obtain an equal number of decision support capabilities that can be fused together to produce the big data decision support information needed by the bid data owner.

The big data owner can either act upon the decision support information on hand or collect more random data subsets. The random data extraction process continue until the big data owner judge a feasible decision can be made.

We use Dempster and Shafer theory and Smets’ transferable belief model to generate feasible pignistic probabilities that the big data owner adopts. The sequential decision model is not treated in this study, but a numerical example is provided to illustrate the generation and the fusion of decision support capabilities.

This proposal will only briefly present an outline of the analytical model we use to produce business value from big data through the generation of specific decision support. The specificity relates to the limited number of decision parameters involved in the decision solving situation. Our intent is to set the mathematically sound framework to employ accepted decision science and related traditional utility theory.