MOOC Learning Behavior Analysis And Dropout Prediction Based On Feature Engineering

Massive Open Online Courses(MOOC)have expanded rapidly since 2012,providing users with a low-cost,high-quality learning experience.High dropout rate is a serious obstacle to the development of MOOC.One of the ways to solve this obstacle is to use the abundant data resources in MOOC to explore the factors of dropout,build a dropout prediction model and establish a warning mechanism.However,due to the huge and complex data of MOOC,these are some problems: loss of data value,mismatch between data and model,and poor replicability of researches.Feature engineering is a data representation method,which has been verified in the field of machine learning.It can be migrated to the field of education,providing a new path for data-driven predictive research of MOOC dropouts.Taking the learning behavior data in the click stream log of MOOC as the research object,this study analyzes the behavior data of MOOC learners and models the prediction of MOOC dropout based on feature engineering.In the study,first of all,the relevant literature of the study is systematically reviewed.On this basis,the core concepts and relevant theories of the study are elaborated,the methods of feature engineering are summarized,and the behavioral data background and dropout prediction model of MOOC learners are discussed.All these lay a foundation for the application of feature engineering to achieve dropout prediction and analysis.Second,taking the learning behavior log of nearly one million learners in 12 Massive open online courses as the data source,the study extracts and processes the data with the method of feature engineering,analyzes the prediction ability of individual behavior features and the correlation between features,and then proposes the schemes of feature selection for the MOOC dropout prediction.Third,on the basis of designing the prediction model of MOOC dropout based on feature engineering,the study uses Logical Regression and Long Short-Term Memory to achieve the prediction of dropout with high accuracy and stability.Finally,combined with the research conclusions,the study discusses the application of feature engineering in MOOC learning behavior analysis,and discusses the implementation and application strategy of MOOC dropout prediction.There are some innovations in the study.First,the study is based on fine-grained behavioral data features to carry out the research of MOOC behavioral feature extraction and dropout prediction.Second,the transfer of feature engineering to the field of education provides a new perspective for MOOC learning behavior analysis.Third,combined with feature engineering and deep learning methods,the study predicts dropping out in advance based on one course data,and reuses into multiple courses to improve the replicability of predictive analysis research.This study aims to solve educational problems from a data-driven perspective and provide a reference for predictive analysis research in the context of MOOC.