The models & results

As a part of a group project, I analyzed how well students learned during the quarantine period due to Covid-19 using machine learning models. I looked at what students did in order to learn effectively and what students could do to improve their learning experience. In this blog post, I will go in-depth about the results of our project and the machine learning models used to produce them.

As mentioned in the previous post, the dataset included 40 features. Although having many features seems like it would yield more accurate results, it would actually result in overfitting: errors in predictions caused by too many features. To find out which features showed the most correlation, our group used a logistic regression machine learning model to find the most meaningful predictors. These predictors, 9 in total, include the learning environment, resources, total learning hours, learning hours before COVID-19, and more. This means that our main machine learning model will now be more accurate since our features are narrowed down.

Our main machine learning model was the K-nearest neighbors (KNN) model. In short, this model will give an outcome based on inputs that are most similar to each other. Here is a visual to help explain.

If you were to add a new dot to either of these plots, its color would be the color of the dots that are closest to it. For example, if you were to add a dot towards the bottom right, it would become the color blue.

This is what this model would look like in terms of our project…

The results of our KNN machine learning model were fairly good. It was able to predict how well a student would learn given certain circumstances with an accuracy of 84.9%.

So what does this model tell us about the changes that should be made in order to improve the learning experience for students? In order for students to have an effective learning environment, they should dedicate more time to learning, both during class and outside of class, have many learning resources, and have an encouraging learning environment. Although some of these factors may not be things that students can control, it’s important to remember that it's also the school’s to ensure that students learn in an effective manner.

In my next post, I will talk about how you can create your own project similar to this one.

Ryan Tietjen is a Student Ambassador in the Inspirit AI Student Ambassadors Program. Inspirit AI is a pre-collegiate enrichment program that exposes curious high school students globally to AI through live online classes. Learn more at https://www.inspiritai.com/.