Design And Implementation Of Analysis And Monitoring Platform Based On Multi-time Dimensional Exercise Data

Nowadays,people pay more and more attention to physical health,and various wearable devices have collected some user-related health and exercise data.However,these data are currently more stored and used for some statistics.These exercise data contain potential information that can obtain personal physical characteristics and movement.If we can use data mining technology to analyze these data,we can mine this potential information and with setting abnormal rules,we can provide users with monitoring services for personal exercise and health.This topic aims to design and implement a analysis and monitoring platform based on multi-time dimension exercise data,and apply data mining technology to analyze the currently available sports-related exercise data.This includes algorithms to separate different exercise styles,as well as algorithms for analyzing and predicting the user's future exercise situation.Compared with some traditional platforms for collecting health and exercise data,this platform not only can collect user data,but also has the function of analyzing user data and monitoring abnormalities of user exercise data.And this platform has a more beautiful and friendly user interaction interface.In order to implement this platform,this thesis first elaborates the research background and significance of this platform,clarifies the main work to be completed and designs the structure of the thesis.Then,this thesis researches,studies and introduces the key data analysis technologies of this platform.Then this thesis clarifies the specific functional requirements of the platform,and elaborates the key issues that need to be resolved in the main functional requirements.The thesis first proposes a detection rule for abnormal data as a preprocessing method and analyzes it on the exercise data set.Then,a clustering algorithm based on Fourier transform is proposed to separate the user's exercise style and a custom search strategy is used to determine the number of clustering clusters.The experimental results prove that this algorithm is more effective than the original K-means clustering method in exercise dataset.Then,a hybrid model based on binary multi-level neural network and support vector machine named BNNS is proposed to analyze the classified historical exercise data in multiple time dimensions and make predictions for future exercise trends.The experimental results prove that the model not only can output the prediction values in multiple time dimensions,but also has higher accuracy than a single algorithm model,and found that the weather characteristics have a great impact on the exercise situation in the analysis of typical user.Therefore,this thesis also uses weather characteristics to make correction predictions of future exercise trends,and uses the shapley value-based feature importance assessment method to visually display and analyze the correlation between weather and air quality characteristics and exercise data.On the basis of solving the key problems,the overall architecture of the entire platform and the interface of each module were designed,and the core system-level interaction process was clarified.Based on this,several core modules of the analysis and monitoring platform are designed in detail,including the design of class diagrams,the design of methods,and the design of the internal interaction process of the module.After the detailed design is completed,the system test conditions are described in detail,including the deployment of the test environment,unit tests,and integration tests.Finally,it points out the further work that this platform needs to complete in the future.