One of the understanding analytics in
MOOC is to determine and forecast students'
performance based on different antecedents that
gathered as records from the system for the engagement
activities of the students. Along with the visibility of big
data, the usage of artificial intelligent methods can
easily supply effective results in forecasting the
students' standing and performance. This study aims to
provide an artificial neural network design for
forecasting students pass/fail status along with their
band performance based on MOOC big data analysis.
The data collection utilized is the one collected and
discharged through Harvard and MIT
"HarvardXMITx -Course Dataset AY2013" in May,
2014. MATLAB Convolutional Neural Networks (CNN)
is used as a platform for simulating the proposed
design. For the data set, the total cases were 641138
cases; the filtered cases with complete field were 58453
cases. The initial design has eight possible input
variables, which is tested as a preliminary step to
determine its importance to the model. The final design
for predicting learners’ performance and level have
four inputs and two outputs. Predicting accuracy of
success status (pass/fail) shows that 91.6% of learners’
success status can be predicted by using testing data.
Predicting accuracy of success Level (Band 1 to Band 5)
shows that 82.6% of learners’ success status can be
predicted by using testing data. The proposed data
mining model has four input variables and the
precedence for its importance are day’s activity,
followed by played videos, then events number, and
finally chapters opened.
Keywords : MOOC, Data Analytics, CNN, Student Performance, Educational Data Science.