Research On Probabilistic Databases And Efficient Query Technology

There are lots of uncertain data in the information retrieval, sensor data and image processing. Database need to process this uncertain information when it is stored in database. The traditional database can not do it very well, so research on probabilistic database becomes more and more important.The paper firstly describes the research background, status, the wider use of the probabilistic database and introduces two extensive applied probabilistic database models. We analyze current shortcomings about probabilistic database and make some improvement about it. Its tuples is re-classified by some standard and taken different formula to calculate the probability. The problem about unreasonable projection is effectively resolved in this method.Fuzzy query technology on figure attribute in probabilistic database is our one main researching. We implement query process from fuzziness to exactness through creating fuzzy sets, choosing membership function and adjusting the fuzzy scale about fuzzy object. Basic concept about confidence is proposed in this paper. By setting efficient confidence, the number of tuples in low probability can be effectively reduced. It will not only in a better way to meet user's requirements, but also can reduce the query time.At present, aggregation function is directly applied into each possible world, but it can not be calculated in linear time. The paper directly applies the aggregation function to the original probabilistic relation and computation is made against each tuple by transformation and store procedure methods. Theory analysis and experiment result proves its correctness. Each aggregation function is divided into three components and it can meet user's various needs in a better way. We propose approximate computation method to calculate AVG. The experiment proves that it can shorten running time and error rate in this way is very low, so we can take it as our exact AVG.