Predicting Student Success
A final project for my Machine Learning and Data Science. I used grades and various background info on students to create a model that predicts whether a student would be successful in Algebra 2.Overview
This was my final project for my Machine Learning and Data Science classes.
I chose this project because it addressed a very real problem I had when I was a math
teacher: will this student be successful in Algebra 2? This may seem like a trivial
question, but in Hawai'i public schools, Algebra 2 is the first "elective" math course that
students take. Historically, it also has a pretty high fail rate.
I would like to point out early on that this is in NO WAY meant to be a replacement for a teacher's opinion, because it is well established that teachers generally know their students best and most could easily answer this question for you. However, we live in a world where everybody wants to see the "data" to back up these opinions. I used grades and various background data on students to create a model that predicts whether a student would be successful in Algebra 2.
Fortunately, I was able to get data from the high school I used to teach at. This was a great opportunity for me to work with real data and go through the entire pipeline of cleaning & processing the data to training models on it. It was not a large enough data set to do any neural networks, but I was able to train a fairly accurate decision tree. One of the most interesting observations from the data was that a student's overall GPA had the highest effect on their probability of success. This may seem like an obvious statement, but in my model overall GPA had a higher weight than their "Math GPA" or Algebra 1 grade. Thus inferring that its not just whether a student is "good at math," but whether they are an overall successful student that determines their success in Algebra 2.
I gave my model back to the math department for them to use with future students with the hope that it would be a useful tool for them to "show the math" to back their decisions. If you would like to know more details about the parameters of my project, please feel free to read my paper on it.
I would like to point out early on that this is in NO WAY meant to be a replacement for a teacher's opinion, because it is well established that teachers generally know their students best and most could easily answer this question for you. However, we live in a world where everybody wants to see the "data" to back up these opinions. I used grades and various background data on students to create a model that predicts whether a student would be successful in Algebra 2.
Fortunately, I was able to get data from the high school I used to teach at. This was a great opportunity for me to work with real data and go through the entire pipeline of cleaning & processing the data to training models on it. It was not a large enough data set to do any neural networks, but I was able to train a fairly accurate decision tree. One of the most interesting observations from the data was that a student's overall GPA had the highest effect on their probability of success. This may seem like an obvious statement, but in my model overall GPA had a higher weight than their "Math GPA" or Algebra 1 grade. Thus inferring that its not just whether a student is "good at math," but whether they are an overall successful student that determines their success in Algebra 2.
I gave my model back to the math department for them to use with future students with the hope that it would be a useful tool for them to "show the math" to back their decisions. If you would like to know more details about the parameters of my project, please feel free to read my paper on it.