Predicting Success in Online Education

Predicting Success in Online EducationA Random Forest Classification ModelJeremy MillerBlockedUnblockFollowFollowingDec 5IntroductionWith the goal of being able to predict whether or not a student will successfully complete their online course, I built a Random Forest Classification model based on the data from the Open University in the United Kingdom..Identifying students at risk for failing or dropping is the first step towards interventions which can lead to greater student success.The goal of this project is to use behavior and demographics to predict whether students will successfully complete the course and to flag students for intervention.ResultsAfter splitting the data based on time and using the first 1/4 of the course time, a random forest classifier provides a true positive rate of ~0.725..(A double-module is worth twice the credit of a single course and scored on a scale of 200.) I then joined and aggregated the data to create features, some of which were:* Demographics and Student Information* Gender* Region* Highest education* Socio-economic status* Age group* Disability status* Student interactions with the course* Average days assessments submitted earlyThe target for my model was:* Course not completedModeling processMy primary modeling concerns were evaluating the classifier based on the true positive rate (recall) and ROC AUC score..I chose the Random Forest Classifier as my final model given its strong performance and relatively high level of interpretability.EvaluationThe final classifier was evaluated using the ROC AUC score and the true positive rate described above and by comparing the model to a relevant baseline.. More details

Leave a Reply