Due to swift development of data mining as well as machine-learning technology and the flare- up of big sports data mining expansion challenges, sports data mining cannot merely use data statistical methods such as how to club machine learning and data mining technology for efficient mining and analysis of the sports data, to supply useful advice for the public physical exercise, and this is an vital need to study. It is a kind of effective sports data mining work through feature selection algorithm. Around the tricky problems existing in the study of the sports effect, given the drawback of existing data sets and conventional research methods, this paper begins from data mining algorithm, construct the sports effect evaluation database, based on the feature selection scheme, using elastic system network algorithm, random forest algorithm, and the impact of sports on the outcome of physical gauges. The evaluation algorithm presents machine learning techniques and the feature selection algorithm to guide sports effect evaluation research. When studying this evaluation problem of the sports effect, according to created sports effect evaluation database, elastic system algorithm is appended to regularize, realize and optimize the feature selection. When selecting features of different sports skills using the information gains marked to rank the significance of characteristics, which can systematically and accurately provide the influence degree of the sports on diverse physical indicators, bring the physical fitness research little more scientific, and can uncover the effect of the sports as much as possible. Experimental results demonstrate that the selected features as well as ground-truth both have good accuracy and good evaluation as match up to the baseline method.