Application Of Constraint Based Rough Set Parallel Algorithm In Enterprise Production Safety Early Warning

With the economy’s rapid development, safety accidents of the enterprise production have occurred frequently, which caused significant loss to people’s life and property. N ipping risk in the bud would be an urgent subject that needs to be solved as soon as possible. Mass data of safety production must conceal a mapping relationship between safety indexes and safety grade. And rough set theory can find potential knowledge from incomplete information, therefore taking advantage of rough set theory to mine available rules from mass data would be an effective way to avoid or reduce safety accidents.Viewing the trade application above, this thesis puts forward a new algorithm called Constrained-based Rough Set Parallel Algorithm(CRSPA), and applies it to the early warning system of enterprise safety production. The research mainly focuses on two aspects:(1) Propose the CRSPA for the lack of effectiveness in the rules obtained from the traditional rough set approaches and deficiency in dealing with growing data in enterprise production. Firstly, the thesis presents a new definition called target information system of rough set which based on constraint, and introduces the concept of constraint that can be used to compare the advantages and disadvantages of the condition attributes. The introduction of constraint increases practicality and effectivene ss of the rough set theory. Secondly, it analyses parallelism of C RSPA by a strategy of decomposition and combination, and designs detail parallel algorithm with MapReduce framework.(2) Design the safety early warning system. It is a closed- loop adaptive system and includes three stages: training, classification and correction. In the training stage, constructing rough set model according to the parallel algorithm, regarding security index and grade as condition attributes and decision attribute respectively, and imputing them into model and outputs early warning decision rules. In the classification stage, matches the real-time monitoring data with decision rules to get warning category and presents it to users. During the period of correction, evaluates t he accuracy of decision rules and fixes the model when accuracy is lower than threshold.This thesis succeed in realizing the enterprise production safety early warning system by JavaWeb technologies, and evaluating the effect of this application in the pe riod of the system’s trial stage, which conforms that the system has high efficiency and practicability.