Абстрактный

BP neural network-based world men decathlon performance development trend research

Zhendong He, Ruimin Hu

The paper makes statistics of world men decathlon world annual best performances in ten years during 2004~2013, applies principal component analysis and BP neural network combinative method to establish prediction model. The paper carries on principal component analysis of statistical data, and solves four principal components. On the basis of principal component analysis result, take previous year men decathlon four principal components as input, and the second year men decathlon performances as output, it establishes BP neural network model. The paper takes neural network training with performances during 2004~2011, uses performances from 2011 to 2013 to detect established prediction model accuracy, applies 2013 principal component data into predicting, and gets men decathlon annual best performance in 2014 is 8750.5 points. Take principal component analysis’ data handling result as input, it solves men decathlon individual event many data and each individual event performance stronger correlations bad impacts on prediction model’s prediction accuracy, enhances prediction model’s prediction accuracy, and expands BP neural network predication model application.