The Application Of Neural Networks Based On Particle Swarm Optimization In Drugs Management

Data Mining is to discover the potential, interesting relationships and characteristics in database. The objects of Knowledge Discovery in Database include: 1) abstracting the interesting patterns and characteristics; 2) demonstrating the order of the database abstractly; 3) recombining the database according to its data semantic to improve its performance. Nowadays, there are various kinds of Data Mining algorithm that can be applied to different fields. In the health aspect, the hospital information system's development was already mature, and accumulated the massive data, these unfinished data relations are quite complex, the misalignment degree is quite high, very difficult to instruct doctor to work directly. Since the beginning of Neural Network, it has been used wildly in the fields of pattern recognition, automatic control, supplementary strategy, signal process, artificial intelligence, etc. Recently, because of its intrinsic distributed memory and fast and parallel computing capability, artificial neural network has become one kind of important methods in Data Mining. The neural network algorithm is mainly used as the rule of classification and to predict problem in data mining; it also can be applied to pattern clustering. This paper thoroughly research and the analysis neural network algorithm, and has made the improvement, applies in the forecast question, has obtained the very good result.In order to apply neural network to the field of data mining, two key problems must be solved well; the first is to reduce the training time, and the second is that the result of data mining should be understandable. At present, BP pattern is most common neural network pattern in data mining. However, BP pattern has the following disadvantages, such as slow convergence rate, falling into local minimum, etc. In order to overcome those disadvantages, I adopt the method of Particle Swarm Optimization to optimize BP algorithm. A new particle swarm algorithm with dynamically changing inertia weight (DCW) is presented to solve the problem that the linearly decreasing weight (LDW) of the particle swarm algorithm cannot adapt to the complex and nonlinear optimization process. The evolution speed factor and aggregation degree factor of t he swarm are introduced in t his new algorithm and the weight is formulated as f unction of these two factor s according to their impact on t he search performance of the swarm. In each iteration process, the weight is changed dynamically based on t he current evolution speed factor and aggregation degree factor, which provides t he algorithm with effective dynamic adaptability, and improve the performance of Particle Swarm Optimization. And based on this method, used the data from HIS, the paper has developed the medicine predict system. In the establishment drugs forecast that above the system has carried on this algorithm realization, passes through the theory and the experiment contrastive analysis, this algorithm enhanced the forecast performance greatly, has certain guiding sense to the hospital drug control.