OPTIMIZING DEVICE HAVING NEURAL NETWORK EVALUATING DEVICE

PROBLEM TO BE SOLVED: To estimate selectively and one call cost data by dynamically applying a neural network to a database management system based on generation selectivity value in order to determine an optimal query search sequence. SOLUTION: This device includes a feature vector extraction element which is connected to data base tables 120 and 122 which convert an input-output parameters of a related user defined routines(UDRs) into base representation that has standard data type. A neural network receives a feature vector output and produces an estimation optimization value that is used by an optimizing device 102. The neural network regularly uses a query that is randomly produced to conduct training or dynamically trainined by taking in data that are generated in the query. Under a prediction mode, the device 102 dynamically calls the neural network in order to produce estimation that is used to determine optimum DBMS database management system 100 query search sequence to the DBMS.