PERC 2019 Abstract Detail Page
Previous Page | New Search | Browse All
| Abstract Title: | Using Machine Learning to Classify Descriptions of Problem Solving Strategies |
|---|---|
| Abstract: | We report on the use of a generative machine learning model to predict the learning outcomes of individual students on a particular task. Students were presented with a problem isomorphic to the "ballistic pendulum", as part of an online quiz that was completed in lab. As a first step to their solution they were asked to describe in words their strategy for solving the problem. Student responses (N>1000) were homogenized to make equations usable in the model and pre-processed to remove noise (non-essential words, spelling errors, punctuation, etc...) then codified using the "bag of words" model wherein each learner's response is represented as a vector whose dimensionality is the number of unique words in the corpus. This text data, along with the actual outcome (correctness) was used to train the naïve Bayesian classifier. We will describe the process used and the results of our approach. |
| Abstract Type: | Contributed Poster Presentation |
| Session Time: | Poster Session III |
| Poster Number: | C35 |
Author/Organizer Information | |
| Primary Contact: |
Jeremy Munsell Purdue University 3205 Edison Dr. West Lafayette, IN 47906 Phone: 765-607-3517 |
| Co-Author(s) and Co-Presenter(s) |
Tianlong Zu, Purdue University N. Sanjay Rebello, Purdue University |
