As TAs and tutors in the CCI department at our university, each of us see firsthand everyday how small weaknesses in a student’s understanding can hold them back from future learning. It’s often difficult and time-consuming to figure out what knowledge gaps may be holding students back from progressing in their learning. We realized that, because tests are often taken online now, a rich store of data that could potentially show us what students are struggling with is constantly being recorded, but it’s being used simply to grade and then it’s thrown away. We wanted to create this tool to make it easier for educators to fully utilize this quiz data to understand how every individual in their class is progressing and how they can potentially improve their students’ class experience.

What it does

We implemented neural networks and machine learning algorithms to create an artificial representation of a student based on anonymized data

How we built it

We built this tool with data from quizzes and tests from multiple sections of a course for the past couple years as a backbone to our coding logic. This data was filtered for specific keywords with an algorithm using the Azure language processing tool. Using neural networks, each questions had a couple keywords that were processed and matched to two factors: accuracy and preciseness represented by a percentage. This was parsed into json which would be readily available to display a real-time data visualization with React, Node.js as students finish assignments.

Challenges we ran into

There was a more steep learning curve to machine learning and data formatting than we had anticipated. A good portion of our time formatting our data and trying to grasp a better understanding of machine learning algorithms and their applications.

Accomplishments that we're proud of

We were able to identify which machine learning algorithms would be the most appropriate for our app’s intended functionalities. We attempted to implement them by making our own algorithms on Python, a language we were not too familiar with, and thought came out with a solid concept that we would like to keep developing in the future

What we learned

Linear Regression, Tree Based Grouping, and their applications; Agile development in Python, Github.

What's next for Iole

We would like to eventually have this become a widespread resource open-source project (MIT Licence) for all educators and the community to improve education as a whole. We would like to finish out and improve the design of our D3/Node.js/React representation and continue working upon the project.

Share this project: