-
Student Stats Vizualization
-
StudentCheck Logo
Inspiration
I come from a family of school teachers in India and experienced first hand the challenges they faced in transition to online learning last year. Most of the school teachers are middle aged and and do not always have the technical skills and cognitive abilities to track the progress of individual students in their class. This problem is further exacerbated in smaller children who themselves struggle with the online learning environment. Further, it is difficult for a teacher to process the past performances of each individual in their class and their previous classes. This inspired us to build a tool which enables teachers to identify the students lagging behind their own potential.
What it does
StudentCheck provides a simple dashboard which utilizes past academic performance of the students and machine learning models to predict when a student is lagging behind their potential. It simplifies the task of identification, making use of student potential rather than absolute ranking within the class.
For every test, we have a set of predicted scores for each individual student based on their past performances. We compare this prediction with their actual performance. We believe that difference in predicted and actual scores reflect unfulfilled potential of the student. Our dashboard makes it plainly clear for the teachers, which students are not fulfilling their potential and where special attention may be best placed.
Evidence from behavioural science suggests that the timing of interventions can be crucial to their success. Identifying when a student lags behind potential, may be a good time for teachers to intervene (rather than waiting for a child to drop further down the class rankings).
How we built it
We utilised Javascript and CSS to create a frontend of the Dashboard. We made use of the PostgreSQL database services on Google Cloud Platform and a backend written in Flask for Python. In addition, our machine learning model was written in Python using the sklearn library.
In absence of a real dataset, we generated synthetic data to mimic shocks in test scores which can be indicative of underlying challenges to latent learning ability.
Challenges we ran into
Participating in our first hackathons, we faced challenges in structuring the project and decoupling the tasks for each individual. There was also a huge learning curve to make the frontend and backend systems as novices in web development.
Understanding the the GCP structure and the linkages between various services it provides and aspects like authentication took us a while to learn. We had planned on using Auth0 authentication for login of the teachers, but were not able to integrate it with our frontend.
Accomplishments that we're proud of
We believe that we been successful in identifying a challenge which will become very relevant as remote studies become more prevalent. Our approach also highlights the distinction between performing poorly and performing below potential. This, we believe, will help teachers to prevent students from descending into a slippery slope. We also feel proud of our abilities to learn and build a service from scratch in 24 hours!
What we learned
We have learned to work quickly, in a team, across locations. This was a challenge, but towards the end of the day (including writing this now), we feel more and more like a well oiled machine. None of us had previous UI/UX experience, and we learned how to make wireframes from scratch. We spent a lot of time thinking about how to make the dashboard as simple as possible for teachers to use.
What's next for studentcheck.tech
We hope to further fix up the dashboard to make it usable for multiple teachers. Fixes include making it easier for teachers to add in new student data (we currently add new scores from the google cloud console). We would like to train our models on real data. This can be done by accessing student grade datasets, and building a better picture of grade dynamics. After this, we would like to pilot the dashboard with real teachers, inputting real historical data, and for it to be updated throughout a term. We would then obtain feedback from teachers to better understand the use of the product. Finally, we would like to work with educational research bodies, to understand what sort of interventions are suitable for children falling behind their potential, enabling us to build our the intervention section of the dashboard. In addition, our product will help facilitate causal evaluations of educational interventions, which may even contribution to our knowledge of what interventions work in the classroom.
Log in or sign up for Devpost to join the conversation.