A Multi-label Learning Algorithm Combining Regression Kernel Extreme Learning Machine With Association Rules

Multi-label learning has solved the problem of classification of polysemy objects in single label learning,and has become one of the hot research areas in machine learning in recent years.The multi-label learning algorithm based on extreme learning machine(ELM)is an efficient multi label learning method.However,the multi-label learning algorithm based on ELM mostly uses the classification mode in ELM,which restricts the efficiency of ELM algorithm in multi-label problem to a certain extent.In view of the above problems,the main work of this thesis is as follows:(1)This thesis uses regression kernel limit learning machine for the first time to solve the multi-label classification problem,and first proposes the regression kernel limit learning machine(ML-RKELM).The traditional classification ELM consumes high time when processing high dimensional data,and the regression kernel ELM directly maps the feature space to solve the analytic solution of the objective function matrix.Compared with the classification mode,the algorithm has great advantages.(2)Using ELM to solve multi-label learning problem,the algorithm does not take into account the strong correlation information between markers.Therefore,for the problem of strong correlation between tags,the multi-label learning algorithm(MLASRKELM)which combines association rules and regression kernel limit learning machine is proposed for the first time.First,the association rules are analyzed,rule vectors between tags are extracted,and then the multi label regression kernel limit learning machine(ML-RKELM)is proposed.The prediction results are obtained.Finally,if the rule vector is not empty,the final prediction result will be calculated by the rule vector and the prediction result,otherwise the final result will be the prediction result of ML-RKELM.The experimental results of the proposed algorithm in 14 open datum multi label data sets show that the two algorithms of ML-ASRKELM and ML-RKELM are better than other algorithms,and the statistical hypothesis test further illustrates the effectiveness of the proposed algorithm.