Large-scale Data Mining Based On Hybrid Evolutionary Algorithms

Feature selection(FS)is an effective data mining method.It can solve the dimension disaster caused by data redundancy by selecting a group of features with high correlation and low redundancy in high-dimensional data.At present,many computational methods have been applied to solve FS problems.Among them,the feature selection model based on teaching and learning based optimization algorithm(TLBO)has attracted more and more scholars' attention because of its efficient global search ability.However,with the continuous expansion of data scale,these algorithms show some limitations,such as insufficient stability,low accuracy and poor local search ability,which leads to the gradual dilemma of algorithm research.In order to solve the above problems,this paper proposes a hybrid evolutionary wrapper algorithm model(TLBOLS)integrating teaching and learning optimization algorithm and local search(LS).Firstly,because the traditional teaching and learning optimization algorithm can not be directly used to solve the feature selection problem,the algorithm converts the real number coding into binary coding in the initialization stage.Then,in order to ensure the diversity of the population,the worst individual restart mechanism is introduced in the teaching stage,and a binary teaching and learning feature selection algorithm(BTLBOLS)is proposed according to the different TF values of learners and teachers in the process of evolutionary class.Then,a local search method combined with multi operation and variable neighborhood search method are proposed to gradually enhance the disturbance and improve the individual quality of the population.In order to optimize the feature selection results,btlbols uses the comprehensive evaluation index as the objective function to guide the overall evolution process.Finally,in order to further solve the problem that multiple objectives conflict with each other in reality,the single objective algorithm is improved to a multi-objective algorithm to deeply solve the problem that multiple objectives need to be optimized at the same time in real life.