An Application of EDM: Design of a New Online System for Correcting Exam Paper

With the new era in information technology, Educational Informationization not only profoundly affects the modern education paradigm, but also provides strong technological feasibility to facilitate and realize education reform. However, the ubiquitous accessibility brought about by today's internet and wireless technologies creates explosive increase in educational related data (big data) from a diversity of resources. It also becomes much more challenging to analyze and understand the underlying meaning of the big data. It is high time that the modern education systems should embed data mining into education paradigm and apply decision analysis to guide teaching strategy. It is our belief that Education Data Mining (EDM) is vital to achieve such educational innovation. This paper proposes a novel Online System for Correcting and Analyzing Exam Paper (OSCAEP) based on a combination of EDM and Network Informatization, which is aimed to provide in-depth analysis of exam papers, to construct an innovative EDM model and to study the interrelationship between educational factors. (1) First, we introduce the background, significance, and purpose of our research. We illustrate OSCAEP from three perspectives: $\pmb{a}$. Educators: For educators, they can get an insight into potential problems from the quantitative results by data mining; $\pmb{b}$. Learners: For learners, they can understand their learning progress from various angles; $\pmb{c}$. Administrators: For administrators, they can timely redesign the education strategies based on such quantitative feedbacks. (2) Second, we present the innovations of OSCAEP from theoretical part and technological part. OSCAEP also include some advanced teaching theories such as Personalized Learning, Precision Teaching and Customized Teaching. Our main technological innovation hinges upon integrating several machine learning data analysis techniques (e.g., K-means, DNN) to train scores and obtain the classification rules. (3) Third, Our presentation of OSCAEP contains two major parts: $\pmb{a}$. build a distributed multi-layer hardware architecture, which includes web server, application server cluster and management workstation, and $\pmb{b}$. provide a multi-stage pipeline software model, which expands the traditional process of data mining into six stages (Data Preparation, Data Filtration, Data Pre-processing, Data Conversion, Model Building and Data Training). From the application perspective, we develop an effective model for examination analysis and learning behavior prediction based on education big data. Examples include statistical score data of different types of questions (Blank, True/False, Choice and Coding). (4) Forth, OSCAEP also offers educators some potentially very valuable suggestions and advices. It is found that the learning performance has no direct relation with the learning time, but it has much to do with the learning habits. With the EDM analysis results, educators can better customize teaching and design efficient rules and regulations in accordance with learner's features. This leads to precision teaching and plays a vital role for a more desirable atmosphere for learning and teaching. In short, OSCAEP processes information in a digital, automatic, intelligent and networked way. As such, it provides more precise and timely information than the traditional approach.