The Design And Study Of A-FCM Algorithm Model

Fuzzy C-Means algorithm is a based on objective function is the vague and class methods and the result is dependent on the clustering of a parameter previous experience knowledge (clustering center and clustering number). At present, FCM algorithm is used to numerous fields:model identification, data mining, vague control, image processing and images cut, vector quantizing and fuzzy logic and so on. However, as most of FCM algorithm and improved algorithms which initial clustering centers and clustering numbers almost are given at random and need many tests will get better clustering results. Therefore, give the reasonable clustering centers and clustering number is very important. With the computer science and technology application and development, the vague clustering algorithm which is based on objective function (FCM algorithm) becomes a new study hot.After the domestic and foreign scholar's effort for many years, FCM algorithm has had a great improvement. But, so far the algorithm's some problems is still not get a better solution and advantage of the algorithm can't get a full display. Some specific improvement measures are proposed for the traditional FCM algorithm not only to enhance the efficiency of the algorithm, but could create many positive impacts on the experiment process and the result.In this paper, for the inadequate of FCM algorithm and how to improve the efficiency of the method executive, we modified the traditional FCM algorithm and proposed an advanced fuzzy clustering means algrothm (A-FCM), major research work and innovation of the paper in the following two aspacts:(1)In this paper, the methods which select the right index have been improved for traditional FCM algorithm, that is, to calculate right value according to the practical problems. Because of the right Index m to FCM algorithm clustering analysis have a great influence, in this way we not only can gain more reasonable m value, but can reduce the errors because of artificial choose right values. Experiments prove, through the use of the m value which is gained by caculated, its experimental result is more ideal.(2) The amendment of degree of membership in FCM algorithm, through the amendment degree of membership constantly which convergence speed can be raised and impact of the classification effect of clustering, reduce the time of algorithm implementation. This can improve the overall performance of the algorithm.In addition, FCM algorithm has itself merits. Especially, when large information be processed, it's the advantage is more obvious. At the same time, in some famous science caculate softwares (as matlab) also contains FCM algorithm, these offer great convenience for the research work.Finally, in matlab simulation tool use traditional FCM algorithm and use A-FCM algorithms get the experimental results to compare, simulation results prove, this text proposed A-FCM algorithms can be more effectively improved performance and efficiency of the method, the results alse be more reasonable.