A Research On A Classification Model Based On BP-Network With Fuzzy Output

Now we enter the era of IDSS.which is based on AI,especially on ANN(Artificial Neural Networks).In the 1980's,W.H. Inmon brought about a new concept of data-warehouse,which got its rapid development and challenged the traditional data mining techniques with an unimaginably large amount of data.In our work here,a model of data mining was developed,which got its foundation from Artificial Neural Networks.hi fact,this kind of model might be called an infant protocol for DSS (Decision Support System),which accepts final users' raw data, integrates various sources of data,cleanses them from garbages,then normalizes them into an intermediatary data file.From this temporary data file,the DSS model can now make a conclusion,which is a fuzzy set.To do this,we introduce a kind of ANN models called BP network,which classifies the records from the normalized data file.In the model.we put our main emphasis on BP network and its training algorithm.ANN has been a branch of AI,and it has some kind of intelligence of human beings,as to memorization and reasoning. Based on the intelligence.we can first train the BP network using the standard records from the final users,then,the network can process the raw data from various sources,and reach its own conclusions .Not as normally the ANN systems do,our model outputs a fuzzy set to act as the final DSS result.During the whole procedure,the core lies in the BP network about the model's efficiency and its effectiveness.As a rule,a commonly multipurposed solution falls behind the specially single-purposed one in its efficiency and effectiveness,so much of work has been put into the try to improve the whole ststem performance,which was largely dependent on the special attributes of the data from the definite final user. You'll see how we have achieved the goals.As to programming language to use to materialize the model concepts,we prefer Java much more to others,such as C/C++,etc.There are several reasons for-in-the preference:first,Java provides us with wider range of application of the model;secondly, Java is in the lead position in security programming with its advanced techniques of cryptograghy and easy-to-use classes,which can help us realize our dream of a secure system.It must be pointed out that, despite the model's capability to deal correctly with the raw data from final users,the model is just a model,which means it is not a kind of product with a full consideration for final users.For example,the user's interfaces of this model is rather ugly and primitive,partly due to the limited functions of the Java language.Besides,there left much work on the BP network itself for us to do.This is because we haven't got comprehensive grasp of the field of AI and ANN,and there are lots of recent advancements of AI and ANN that remain hidden to us. And these are just the goals for us to strive for.