Research And Application Of Natural Computation Method Based On Cosine Similarity

With the advent of the information age,scientific and technological research has developed rapidly at an unprecedented speed.By simulating the way of life of natural creatures,the natural computation method has powerful complex information processing ability,wide application range,strong robustness and is loved by scholars.However,the traditional natural computation method also has some problems,that is,the convergence speed of the algorithm is slow in the later stage,the diversity of the population is reduced,and it can not escape when falling into local optimization,resulting in the unsatisfactory optimization results.In order to improve the performance of the algorithm,a variety of optimization strategies have been proposed.Based on the natural computation method,this paper studies the similarity between population individuals,and proposes an optimization strategy in the iterative process of individual updating.By calculating the cosine similarity between the individuals of the population,the individual optimization ability is improved,balance the exploration and development functions of the algorithm to improve the accuracy and effectiveness of algorithm optimization.Then Markov chains are used to analyze and test the convergence performance of the algorithm.The time performance of the improved algorithm is also calculated.In addition,taking the BP neural network as the research object.Finally,the effectiveness of the improved model is verified by real estate price prediction experiment.The main research contents and results are as below:A natural computation method based on inverse cosine similarity is proposed.The strategy adequately considers the reverse in the specific iterative process of the population,calculates the similarity between each individual and the regional central individual,and classifies the population according to its cosine attribute value;The individuals with low similarity after classification are used to calculate the similarity in the regional center,and then the cosine attribute value is used as the weighting coefficient to reverse calculate the individuals to obtain a new population individual.Speed up convergence and reduce running time.Finally,Cauchy perturbation is carried out on the optimal individual of the population to increase the variation range and thus improve the population diversity.Through comparative experiments,the improved particle swarm optimization algorithm and the simple improved gray wolf optimization algorithm are selected to embedded in the parameter initialization of BP neural network.Finally,the optimized BP neural network model is applied to real estate price prediction of second-hand houses in Beijing to verify its improvement effect.Experiments indicate that the whole performance of the BP neural network has been greatly enhanced.