Research On Landslide Displacement Prediction Based On Intuitionistic Fuzzy Particle Swarm Hybrid Optimization

Landslide is one of the major geological disasters in China,which causes huge losses every year.Therefore,landslide displacement prediction has become one of the geological problems to be solved urgently.Influenced by the acquisition accuracy of the sensor and the signal loss during transmission,the landslide data collected have some problems,such as fuzzy and difficult to identify the characteristics.Intuitionistic fuzzy sets have the ability to distinguish and represent uncertain problems,so that they can better describe and express the fuzzy characteristics of landslide data,and provide accurate modeling for subsequent calculations.In addition,there are many influencing factors leading to landslide displacement and the relationship between the factors is complicated.In order to achieve good prediction results,machine learning algorithms must select the optimal combination of these complex influencing factors as training features,so the selection of landslide features is a type of NP problem.The random and distributed characteristics of the swarm intelligence optimization algorithm give it a natural advantage in dealing with this NP problem.Therefore,this research combines intuitionistic fuzzy sets and swarm intelligence optimization algorithms to conduct in-depth research on landslide displacement prediction.The main work is as follows.Perceived fuzzy sets have the ability to distinguish and represent uncertain problems,so that they can better describe and express the fuzzy characteristics of landslide data,and provide accurate modeling for subsequent calculations.In addition,there are many influencing factors leading to landslide displacement and the relationship between the factors is complicated.In order to achieve good prediction results,machine learning algorithms must select the optimal combination of these complex influencing factors as training features,so the selection of landslide features is a type of NP problem.The random and distributed characteristics of the swarm intelligence optimization algorithm give it a natural advantage in dealing with this NP problem.Therefore,this research combines intuitionistic fuzzy sets and swarm intelligence optimization algorithms to conduct in-depth research on landslide displacement prediction.The main work is as follows.(1)Aiming at the problem that particle swarm optimization is difficult to balance global exploration and local development,an IF-Memetic Hybrid double particle swarm optimization(IFMHDPSO)is proposed.The social reinforcement operator and collision rebound operator are designed to improve the population diversity.Intuitionistic fuzzy multi-attribute decision making is established to comprehensively evaluate the exploration region and generate the possible global optimal solution region,which can guide the development population with Lamarckian learning to carry out local fine search.The results of 23 benchmark functions test with 5 new evolutionary algorithms show that the proposed algorithm has better comprehensive optimization ability.(2)Aiming at the precocity convergence problem of particle swarm optimization,multi-topology hierarchical collaborative particle swarm optimization algorithm(MHCHPSO)was proposed.Levy flight is utilized to provide diversity for MHCHPSO.Lamarckian mechanism is used to improve the local fast convergence development of MHCHPSO.The nonlinear decreasing weight with activation and the intuitionistic fuzzy entropy adaptive weight are designed respectively to improve the escape ability of the algorithm after falling into the local optimum.Finally,the multiplicity and convergence of MHCHPSO are adaptive balanced by generation difference analysis and convergence judgment.Experiments on several benchmark test sets show that the proposed algorithm has stronger comprehensive ability.(3)Aiming at the combined explosion problem of landslide feature selection,a reliability-enhanced surrogate-assisted particle swarm optimization(RESAPSO)was proposed.The Bayesian evaluation strategy is proposed based on Bayesian theorem.The form of a posteriori probability of the strategy reflects the degree of fitting between the proxy model and the real objective function,so as to improve the prediction result of the model.Intuitionistic fuzzy multi-attribute decision making is used to evaluate the appropriate individuals to improve the ability of proxy model fitting.A variety of intuitionistic fuzzy information extraction methods are proposed as the decision variables of multi-attribute decision to improve the decision precision of multi-attribute decision.Experimental results on several benchmark test sets show that the proposed algorithm has better comprehensive ability than other advanced surrogate-assisted evolutionary algorithms.Finally,the proposed algorithm is applied to feature selection and super parameter optimization of neural network,and applied to landslide displacement prediction,and the results are superior to the traditional methods.