Architecture for Large Scale Reasoning in Business Intelligence
Xinlong Zhang, Bin Cao, Yanming Ye · 2012
Business intelligence plays a crucial role in modern business.Nevertheless, present business intelligence is not in a position to provide comprehensive business advices owing to limitations on the scope of data and satisfy the indispensable timeliness for business activities.To address these problems, we propose an architecture for business intelligence which could reason on data from numerous domains and provide different users with disparate business advices and results.Furthermore, in our architecture, the production system used to reason depends on MapReduce programming model to implement production rule matching concurrently in different computers with the Rete algorithm.Adopting MapReduce programming model enables production system to obtain more impressive efficiency in rule matching, especially when it comes to a large-scale rules and facts.What's more, we also adopt two conflict-resolving polices to decide in which sequence matched production rules are executed.In this paper, we firstly describe the architecture and then illustrate the particular implementation of this architecture.