Modied Fuzzy Min-Max Neural Network for Clustering and Its Application on the Pipeline Internal Inspection Data

In this paper, an unsupervised learning algorithm called the modied fuzzy min-max neural network for clustering on the application of the pipeline internal inspection data (MFNNC) is proposed. As the original fuzzy min-max clustering algorithm, each cluster of the MFNNC is a hyperbox. And the hyperbox is decided by its membership function. The size of the cluster is determined by its minimum point and maximum point. Compared with FMNN by Simpson(1993), the MFNNC has stronger robustness and higher accuracy, which has proposed an boundary rule and also taken the noise into account. Through the MFNNC, the problem of the points on the contraction boundary has been solved. And the inuence of noise on the whole algorithm is reduced. The performance of the MFNNC is checked by the IRIS data set. The simulation result shows that the MFNNC has better performance than the FMNN. At last, the application on the oil pipeline is given. The result shows that our modied algorithm scheme can be regarded as a method to preprocess for the classication of the pipeline internal inspection data.