Research On Multi-label Data Classification Based On Extreme Learning Machine

With the development of the Internet and big data technology,parties are increasingly dependent on data,resulting in larger and larger data scales and higher complexity of data samples.In order to accurately predict data and obtain accurate and valuable information,it is necessary to select appropriate technologies for processing With the advantage of the Extreme Learning Machine model,this paper carries out related research work on data labeling and data sample structure in multi-label classification.The main work includes:For multiple tag exists in the nonlinear data samples and repetition of sample data,this paper puts forward a kind of extreme learning machine based on the online order samples of the improved algorithm of linear and data preprocessing extreme learning machine The algorithm of linear inseparable data samples by a kernel function,make with linear separable feature data sample For the processed data samples,the online sequential extreme learning machine is used to preprocess the classified data before calculation,that is,different feature labels are found from the training and test data sets and the newly generated tag groups in the experiment are saved,so there will be no repeated feature labels,which greatly reduces the comparison times of training Experimental results show that compared with other extreme learning machine models without sample linearization and data preprocessing,the calculation accuracy is greatly improved,and the calculation time is also reduced.Extreme learning machine for large and complicated labels classification problem,and easy to cause memory problems and will not be able to achieve good generalization accuracy So this article use the stack type extreme learning machine,dedicated to solve large-scale and complicated labels classification problem stack type extreme learning machine learning not only faster And stronger generalization ability Stacked Extreme Learning Machine is approximately to divide a very large Extreme Learning Machine into several very small and series Extreme Learning Machines according to certain provisions and principles In its memory footprint is small and the network under the condition of relatively fixed size,could have high dimensional extreme learning machine learning data in the space,very good to solve the current sample data processing for the huge and complicated data issues the experimental results show that the stack type extreme learning machine compared to other tabbed classification model,predict a shorter time and higher accuracy.