Research Of Fuzzy Clustering Algorithm For Optimizing Incomplete Data Based On Extreme Learning Machine

In real life,the phenomenon of incomplete data collection occurs frequently.There are various factors that leading to incomplete data sets.For example,the data set is incomplete due to the following reasons: acquisition equipment failure,transmission media failure,storage medium failure and detection instruments,and so on.Three traditional methods of treatment: abandon,modeling,filling.It is an important research topic to select an appropriate way to deal with incomplete data sets,and to make the fuzzy clustering analysis.The main work of this paper is based on this.First of all,the traditional fuzzy clustering algorithm deals with the complete data set.In this paper,a fuzzy clustering(FELM-FCM)algorithm is proposed for the optimization of information feedback learning machine.The network parameters of the learning machine are random,the training is simple and the output value is excellent.Therefore,this paper combines the relationship between data and data attributes between the two factors.The complete data set as the network model of the training sample set,combined with Calman filter,the actual network prediction value and the theoretical expectation difference feedback to the output layer,then forecast missing attribute information feedback of extreme learning machine(FELM).All errors by using FELM network model between the training data to obtain the desired output and actual output,and then forecast valuation of the missing attribute,finally realizes the optimization of FELM valuation recovery after the complete data set,and this set of data are analyzed by fuzzy cluster analysis.Secondly,the missing attribute values are fuzzy and uncertain.In the form of points in space,the expansion ability and fault tolerance are poor.Thus,the interval valued fuzzy c-means clustering(FELM-IFCM)algorithm is proposed for the optimization of information feedback learning machine.The use of interval rule is proposed,the FELM network will be the missing attribute value of the interval numerical valuation as,estimation error is obtained as interval width,realize the transformation of the form attribute interval missing,while the complete attribute also expressed as the interval form.Thus,the fuzzy clustering analysis of the transformed interval data sets is carried out.Finally,the simulation experiment is carried out on the MATLAB platform using the artificial data sets and the collected data sets.The experimental results show that the FELM network is used to optimize the incomplete data set,and the accuracy of the clustering result of the restored data set is improved compared with the contrast method.The interval is more accurate and robust than the numerical form.