Data Analytics for education is fast growing into an important part of higher learning institutions, which helps to improve student success rate and decision-making with regards to teaching methods, course selection, and student retention. The undergraduate program at Texas A&M University requires students to take up a general engineering program during their freshman and sophomore years. During the course of this period, student academic performance, abilities and participation is assessed. As per the Entry-to-a-Major policy, departments place the students in the best possible major based on their displayed capacities and in alignment with their goals. Our focus is on the Electrical Engineering department and the success rate of students with aspirations and background in this major. An approach to improve student retention rate is to predict beforehand the performance of students in specific course disciplines based on the information that is mined from their previous records. Based on the outcome, decisions can be made in advance regarding their further enrollment in the area and need for specific attention in certain aspects to get students up to the benchmark. In this thesis, we put together a set attributes related to students in the general program and with an electrical engineering aligned background. The analysis centers around building a method that explains the joint influence of attributes on our target variable and comparison of prediction performances between our models. The prime tools used are Supervised classification and Ensemble learning methods. We also develop a metric-based learning framework suitable for our application that enables competitive accuracy results and efficient pattern recognition from the underlying data.
Data Analytics for education is fast growing into an important part of higher learning institutions, which helps to improve student success rate and decision-making with regards to teaching methods, course selection, and student retention.
The undergraduate program at Texas A&M University requires students to take up a general engineering program during their freshman and sophomore years. During the course of this period, student academic performance, abilities and participation is assessed. As per the Entry-to-a-Major policy, departments place the students in the best possible major based on their displayed capacities and in alignment with their goals. Our focus is on the Electrical Engineering department and the success rate of students with aspirations and background in this major. An approach to improve student retention rate is to predict beforehand the performance of students in specific course disciplines based on the information that is mined from their previous records. Based on the outcome, decisions can be made in advance regarding their further enrollment in the area and need for specific attention in certain aspects to get students up to the benchmark.
In this thesis, we put together a set attributes related to students in the general program and with an electrical engineering aligned background. The analysis centers around building a method that explains the joint influence of attributes on our target variable and comparison of prediction performances between our models. The prime tools used are Supervised classification and Ensemble learning methods. We also develop a metric-based learning framework suitable for our application that enables competitive accuracy results and efficient pattern recognition from the underlying data.
