Application Research Of Distributed Whale Optimization Algorithm

Nowadays,the pace of modernization of our society is getting faster and faster,which has driven the rapid development of the city's economy,and has also greatly improved the overall living standards of our citizens.What has changed is that the population of large cities is constantly expanding,the load of public transportation is increasing,and more and more people are choosing green public transportation.As a result,the city's public transportation is gradually difficult to load the rapidly increasing travel demand,and the urban transportation environment is becoming more and more severe.Handling public transport issues and planning and making decisions on public transport are the key to solving traffic congestion.In the era of big data,scientifically using cloud computing platforms and related technologies to analyze and judge public transportation data to make reasonable predictions and decisions.This thesis proposes a feature selection method of distributed whale optimization algorithm based on Spark platform,improves the distributed whale optimization algorithm,and then combines a random forest prediction model to propose a combined bus passenger flow prediction model.The specific work of this thesis is as follows:(1)The initialization strategy and distributed improvement ideas of Whale Optimization Algorithm(WOA)are studied.Aiming at the way the whale optimization algorithm initializes the population,a strategy of reverse learning initialization is proposed to improve the diversity of the population during initialization and avoid premature convergence.Using binary coding,the whale optimization algorithm is distributed,and a feature selection method based on the distributed whale optimization algorithm is proposed.By comparing the Particle Swarm Optimization Algorithm and the Genetic Algorithm in a stand-alone experiment,it is found that the feature selection method of the whale optimization algorithm has a better optimization effect;through the analysis of distributed experimental results,iterative algorithms such as the whale optimization algorithm,the distributed platform The efficiency in processing big data has improved significantly.(2)For the search process of the whale optimization algorithm,this thesis first proposes the design of a nonlinear convergence factor,which can better balance the performance of global search and local search,and then further solve the problem by introducing a sine and cosine search strategy The algorithm tends to fall into the defect of local optimization when solving large-scale problems,and then the improved algorithm is distributed,and a Spark-Improved Whale Optimization Algorithm(SIWOA)based on Spark is proposed.Combining SIWOA and Random Forest algorithm,a SIWOA-based Random Forest prediction model(SIWOA-Random Forest,SIWOA-RF)is proposed.Finally,through the 12 sets of test function experiments on a single machine,the basic WOA and SIWOA solutions are compared to find the best results.The results show that SIWOA is superior to WOA in terms of single-peak function and multi-peak function solution;then through a city of China The data is used to compare the prediction effect of the SIWOA-RF prediction model and the traditional RF prediction model.The results show that the SIWOA-RF prediction model is superior to the traditional RF prediction model in the final prediction result.In general,improving the whale optimization algorithm through nonlinear convergence factors and sine and cosine search strategies has significantly improved the solution efficiency of the whale optimization algorithm,and distributed the algorithm to use the whale optimization algorithm on the Spark distributed platform Iterative processing of big data can effectively improve the efficiency of the algorithm.