The Uncertainty Estimation Of The Transaction Access Data And Priority Assignment In Real-Time Probabilistic Database Systems

Real-time databases are widely used in the market forecasting, manufacturing control,sensor network, military command system and other fields. These applications couldproduce uncertain data. As the traditional real-time database cannot manage the uncertaindata effectively, it is necessary to integrate real-time database technology and probabilisticdatabase technology for the uncertain data management and real-time concurrentprocessing.In real-time database system, transaction scheduling, as one of the most importantresearch objects, is considered mainly by the urgency and the critical of the transaction.However, in current probabilistic database system, the transactions are scheduled using thetraditional database scheduling, which ignores the real-time of the processing. As the datauncertainty and the real-time requirement of data processing in the real-time probabilisticdatabase system, the priority assignment and scheduling of transactions should considerthe quality of the processing results of transactions, the urgency and the critical at thesame time.It is infeasible to evaluate the quality of the processing results in the transactionpre-analysis process, so we propose a method to estimate the uncertainty of the access dataset of each transaction. However, when the database stores numerous tuples or relations,accessing the data directly would cost a lot of time. In order to enhance the effect of theuncertainty estimation of transaction accessing data, a novel histogram model(P-histogram) to compress the uncertain data is proposed for the rapidly uncertaincalculation. What’s more, an improved method based on the P-histogram model is presentfor the accuracy of the estimated uncertainty. Then we propose the artificial combinationand the nature combination priority assignment policies, and finally, evaluateexperimentally that the policies proposed in this thesis not only maintain the features ofthe traditional policies of real-time transaction, but also can reduce the uncertainty ofresult set more than10%.