International Journal of Innovative Technology and Exploring Engineering (IJITEE) ISSN: 2278-3075, Volume-8, Issue-6S3, April 2019 22 Published By: Blue Eyes Intelligence Engineering & Sciences Publication Retrieval Number: F10050486S319/19©BEIESP Abstract— Applying data mining and machine learning techniques on Moodle logs is an emerging trend that can help track student’s performance and decrease the failure rate. Due to Moodle’s limitation to provide these features, this study was conceptualized. The study made use of historical data from Moodle logs of past academic years to pre-process and develop machine learning models using an open source data mining tool named Weka. This study made use of predictor attributes related to study behavior of students such as Course Viewing Time, Resource Views, Quiz Taken, Replied in Discussions, and Viewed at Weekends. However, it was found out that predictor attributes such as Activities Completed, Course Views and Assignment Passed are the ones which are strongly correlated to students’ performance. Moreover, the predictive accuracy of a model improves depending on the machine learning algorithm being used. Algorithms such as J48, Random Forest, JRip, and OneR have been consistently performing well regardless of the model it is being trained into; and, achieved a predictive accuracy as high as 96.42%. The study was able to reflect the predicted results of Weka back to Moodle through an integrator and developed block using Moodle API. Finally, the developed application was evaluated by IT Experts using the ISO 25010 criteria. Index Terms—data mining, machine learning, predictive analytics, predict students’ performance, Moodle logs I. INTRODUCTION The growth of information available online as well as the stored data in huge organizations led to a quest of discovering hidden information that is useful for decision making. Imagine shopping malls and supermarkets using all the swiped data of their customers to identify frequently bought items and turn this information to targeted marketing and loyalty programs. An emerging field in education called Educational Data Mining (EDM) is designed for automatically extracting meaning from large repositories of data generated by or related to people's learning activities in educational settings [1]. For example, several learning management systems (LMSs) track information such as when each student accessed each learning object, how many times they accessed it, and how many minutes the learning object was displayed on the user's computer screen. LMS accumulate a vast amount of information which is very valuable for analyzing students’ behavior and could create a gold mine of educational data [2]. By applying data mining techniques on student data in LMS, we can obtain knowledge that can help to improve quality of education, student’s performance and decreased failure rate. It is also Revised Manuscript Received on April 12, 2019. Edmund D. Evangelista, DIT Student, St. Paul University of the Philippines, Cagayan 3500, Philippines. (E-mail: edmundevangelista@ymail.com) helpful in early identification of dropouts and students who need special attention and allows the teacher to provide appropriate advising/counseling. LMS produce information of high educational value, but usually so abundant that it is impossible to analyze it manually [3]. Tools to automatically analyze this kind of data are needed. Unfortunately, these platforms do not provide specific tools to allow educators to thoroughly track and assess all learners’ activities while evaluating the structure and contents of the course and its effectiveness in the learning process [4]. Moodle, even in its latest learning analytics can only detect students at risk of dropping but not failing. With these limitations of predictive analytics in the existing framework, this study was conceptualized. This study aimed to develop machine learning models using study behavior predictors of students’ academic performance through Moodle logs. Specifically, historical data from Moodle logs were harvested and analyzed using Weka (Waikato Environment for Knowledge Analysis) data mining tool to determine the study behavior predictors that are correlated to students’ academic performance. Weka is an open source suite of machine learning software written in Java that contains a collection of visualization tools and algorithms for data analysis and predictive modeling. Datasets were harvested based on initial predictors such as Course Views, Course Viewing Time, Resource Views, Quiz Taken, Assignment Submitted, Discussion Views, Replied in Discussions, Viewed in Mobile, Viewed at Night, Viewed at Weekends, and Final Grade. Best features were identified among the initial predictors through the feature selection techniques of Weka and repeatedly trained it using algorithms such as J48, Random Forest, JRip, OneR, Logistic Regression, Multilayer Perceptron, Bagging, and K-Nearest Neighbor. The algorithm which acquired the highest predictive accuracy having false positive and false negative predictions will be used to develop the models. II. CONCEPTUAL FRAMEWORK The conceptual framework of this study was adopted on the concepts of Cross Industry Standard Process for Data Mining (CRISP-DM) and Predictive Analytics Process Model [5]. CRISP-DM, is an open standard process model that describes common approaches used by data mining experts making it the most widely-used analytics model in data mining [6]. It was conceived in 1996 and became a Development of Machine Learning Models using Study Behavior Predictors of Students’ Academic Performance Through Moodle Logs Edmund D. Evangelista