Research On Academic Early Warning Algorithms For College Students Based On Big Data

Carrying out teaching management based on educational big data,building a modern teaching management and service system is an important step to promote the construction of digital campus and data to promote decision-making.In order to provide students with dynamic and personalized academic early warning services,and provide data-supported decision-making to university managers,this article collects early-warning data based on the characteristics of educational data,and builds an academic early-warning mechanism based on data mining and machine learning methods.The use of data mining technology to analyze educational data has become one of the important constructions for improving teaching management and teaching level in universities.However,because the traditional academic early warning model of colleges and universities is not scientific enough based on data,and algorithms are not adapted to data and other factors,resulting in low prediction accuracy and early warning lag.In response to this problem,this article mainly studies and analyzes the characteristics of educational big data,collects early warning data based on multi-source heterogeneous data,uses mathematical statistics,principal component analysis and other methods to select data items,and normalizes the data to an analyzable data set through data preprocessing.A supervised fine-tuning FT_BP neural network algorithm combined with Adam’s algorithm is proposed,and its performance is compared with that of classical classification algorithm.In the end,the behavioral factors,basic information,and historical performance of 3381 school students were collected by experiments.PCA method was used to select features and reduce redundancy.After data preprocessing,classified by FT_BP neural network,established a database to store early warning results and displayed them on the web pages.After testing,the algorithm can effectively improve the problem of low accuracy,can greatly improve the quality of early warning,can predict the academic trend through the development of student learning behavior,and provides school administrators with a scientific basis to guide and manage students.