Predicting Dropout from Online Education based on Neural Networks

Abstract

While online education keeps expanding, web-based institutions face high dropout rate, pushing costs up and making a negative social impact. Based on the analysis of existing research, personal characteristics and learning behavior were selected as input variables to train a dropout prediction model using neural network algorithm. The outcomes of pre- diction model were analyzed by calculating the rates of accuracy, precision, and precision. The results suggest this method is effective in identifying potential dropouts, and can help the online education institutions prevent dropout. As China pushes harder on its industrial transformation and reconstruction, the society has an increasingly high de- mand on the quality of employees. Receiving further educa- tion has become an important way for people to improve their knowledge structure. Online education has been well received for its openness and flexibility among students who are also in the workforce. According to a statistic report on China's higher education, 1.96 million college and universi- ties students were admitted into 68 web-based colleges (in- cluding radio and television universities) around the nation in 2012 (1), taking up 45% of the total number of admitted students into higher educational institutions that year. About 55.7% of people are willing to receive online vocational training, according to statistics from social research institu- tions. Several large enterprises in China have web-based colleges for staff training. Online learning market has become mature, but the prob- lem of high dropout rate is a severe challenge that web-based educational institutions face. Student dropout rate for web- based curricula education in China is generally about 20%, much higher than that of traditional classroom education, according to statistics (2). A high dropout rate is not only a waste of education resources but also a loss for personal education investment. What is more, it leads to a decline in social recognition of e-learning, which is an impediment to the development of web-based education (3). How to effec- tively bring down dropout rate has been an urgent problem of e-learning and attracted massive attention from web-based educational institutions. rate, major of study, etc (4-6). Based on empirical analyses, scholars have explained the factors influencing student drop- out of e-learning through many models they proposed, and have been trying to reduce the dropout rate by avoiding negative factors and improving positive ones in a macro- scope (7). However, as there are huge individual differences among students, a macroscopic strategy is often untargeted and ineffective. A feasible measure to reduce dropout rate is to identify the groups of students to quitting so that targeted measures can be taken before attrition happens. This article adopts data mining technology and predicts student dropout based on related attribute data in the learning management system so as to reduce dropout rates in online education.