Research On Prediction Model Of National Fitness Monitoring Indexes Based On MRMR-IAGA And BOA-Stacking

The current collection methods of national fitness monitoring indexes are restricted by sites,personnel,facilities,data recording methods and costs,etc.,which makes it difficult to form a comprehensive,timely and cheap acquisition mechanism of national fitness monitoring indexes.This dissertation aims to provide an efficient and cheap predictive method for national fitness monitoring.Based on the national fitness monitoring data sets,it aims to establish a national fitness monitoring indexes prediction model.The relevant issues in the field of machine learning such as hybrid feature selection methods,ensemble learning methods and hyperparametric optimization is carried out.Firstly,combining with the characteristics of the national fitness monitoring data sets,from the perspective of reducing model complexity,improving the prediction effect and generalization ability of the model,the study on hybrid feature selection algorithm is carried out,and an algorithm based on max-relevance and min-redundancy and improved adaptive genetic algorithm(m RMR-IAGA)is proposed which achieves the filtering of irrelevant and redundant features and the performance of the algorithm is verified by experiments.Secondly,in order to improving the model prediction effect,through the analysis of the insufficiency of the Stacking ensemble learning algorithm,the study is carried out from the aspects such as base regressor selection algorithm,hyperparameter optimization algorithm,data generation of ensemble model test set and so on.A prediction algorithm based on bayesian optimization algorithm and Stacking ensemble learning(BOA-stacking)is proposed,and the framework of the prediction model is bulit.Finally,on the Windows operating system,Python3.7 is used in the Anaconda3.0environment,based on the m RMR-IAGA hybrid feature selection algorithm and the BOAStacking ensemble learning prediction model to build a national fitness monitoring indexes prediction model.The experimental results show that the model has better prediction effect and generalization ability than other ensemble regression models.