Christie B

I reflected on my project and thought about what I had learnt as well as what I created.
Here is a exert for those interested:

I learnt and revised lots of things while creating this project, including but not limited to:
-> How to create and refine a linear regression model using SciKit-Learn
-> How to find and clean data for machine learning projects
-> How to save a model using Pickle and load it in a backend
-> How to create a simple Flask and JS backend
-> How to create a simple yet effective UI using HTML and CSS
-> How to make a website responsive and accessible
-> How to make a website secure
-> How to deploy an app using PythonAnywhere

The project is not perfect, and there are many ways I could improve it, such as by using a more complex model to lower the MSE, finding better data, and adding more features. I think that the small amount of data made the model less accurate than it could have been, and I also think that there are some features that I could have added that would have made the model more accurate, such as the student’s previous exam scores or their average grade.

This project had a big learning curve, due to the fact that it was my first machine learning project.
However, I learnt to use available resources, both online and otherwise, to efficiently and effectively to find answers to my questions and learn how to do new things.
Some of the most prominent challenges included not knowing how to use new technologies and getting my model to be more accurate.

Thank you everyone for your support!
P.S. I’m saving up for a Polaroid Go Gen 2, so I would really appreciate generous voting. I don’t want to seem greedy or anything, but just so you know, you’ll really make my day by being kind <3

I prepared my app for deployment, adding requirements.txt and setting it up with PythonAnywhere.
It now works at christie.pythonanywhere.app!
I’m really happy that I was able to deploy it so easily and without those hours I spent crying over Nest, however powerful it may be :))
Yayyyyyy!

I added a section that gives people tips on how they can improve their exam score, making sure to make it responsive and accessible.

I made my website mobile responsive by generating some code with GitHub Copilot, and then editing it to suit my needs. It – the AI – basically created a whole bunch of media thingys with a minimum width of 768 pixels. I didn’t really like this, so I consolidated everything into one media thingy so that I can edit the mobile version of the website more easily.
It looks really nice now!
Let me know what you think :)

I added CSS to my website!
I used a light academia theme – a plaid background, and a colour palette cream, beige, and some darker brown accents.
I hope you like it!

I created a frontend, connecting my model to an interface using Flask.
It:

• Jsonifies form data
• Sends this form data to the model api
• The api unpacks the json and creates a prediction from it
• The prediction is jsonified
• The frontend, seeing this, displays the stringified json prediction to the user

I hope to add CSS soon!

To give the website a more distinctive and hancrafted feel, I added a range of custom visual elements to make the website more unique, such as hand-drawn stars, polaroid-style photo frames, and gingham tape accents, contributing to a more playful website vibe.
This will hopefully make the site feel less corporate and more community-driven, reflecting the collaborative and creative spirit of KiwiHacks.
I changed the image of previous hackathons to a corkboard-style collage to add visual interest while still keeping the content organised and easy to engage with.
In addition, I implemented a custom footer that includes the handwritten-style signatures of the KiwiHacks team – reinforcing the personal, human aspect behind the event – and a I added a dedicated sponsors section to acknowledge partners in a structured and visually consistent way.

I created the first draft of the website with HTML and CSS inside of React.
The involved designing, coding, and styling the web page.
A key goal of the project is to move away from a generic, vibe-coded site and instead create something more original – this means that I had to spend abundant time on the design and visuals, making sure that they aligned with the KiwiHacks branding.
The reason why this first devlog is so late is because I didn’t know that I could count this for Flavourtown 🤣🤣

I finished creating and refining my AI model.
I
-> normalised my data
-> added more features

My MAE decreased to 8.3, which I’m happy with for now.
I will save the model with pickle and start working on my backend next.

I:
-> finished importing all required models
-> generated a correlation matrix
-> used online tutorials and ChatGPT to learn how to train and evaluate a model
-> trained and evaluated my model
But the MAE is quite high, at an average of 9 points off when I’d like it to be more precise. I will need to find a way to improve my model.

I:
-> finished planning my project, finalising how I’m going to make it step-by-step
-> found and cleaned up data for my project
-> started installing necessary libraries for my project.

I’ve run into some issues I hope to fix later.

I continued planning my project.
I finalised:
-> what it will do (predict exam scores and give tips to help students based on their input)
-> what technologies it will use (Python/Flask, HTML/CSS, scikit-learn, etc.)
-> it’s structure

I started planning my project.
This involved defining my prior knowledge and interests to hone in on a specific goal – what do I actually want it to do? What am I able to do? What do I want to create?