GhostB

Inspiration

Since many people, including myself, struggle with saving money such as spending impulsively on non-essential items, we decided to create a feature to combat this problem. The idea behind Ghost Budget is to turn saving into a subconscious habit rather than a conscious effort. This project subtly alters spending behavior without requiring users to make tough decisions in the moment. By secretly inflating prices on unnecessary purchases, Ghost Budget helps users save without feeling like they’re making sacrifices.

What it does

This project is a website that uses models and machine learning to track users' spending habitats and secretly inflates the price of their most "unnecessary" purchases in order to reduce the cost of their spending. The extra money that is saved is hidden until the end of month where the user then has the option to claim the money or continue saving.

How we built it

We built this website by first designing the frontend by using languages like HTML, CSS, and JavaScript . Then, we built a machine learning model using over 500,000 bank transactions on Google Collab. We used SQLite for building a database that is more efficient for demo. Finally, we used FastAPI to connect between the machine learning model and the database to process the data from the frontend and display that processed data.

Challenges we ran into

Some challenges we ran into was when setting up the machine learning model. The original dataset size was 22 GB so we had to break it down to smaller pieces to train the model. We faced problems while trying to process that data such as duplicating columns while trying to filter data. The model was biased during the first half of training which resulted in a 1.0 accuracy. The machine learning was recognizing but not regenerating.

Accomplishments that we're proud of

This is the first time we trained a machine learning model and did a project that is based on data analysis. We also finished much earlier than expected compared to the last hackathon we went to.

What we learned

We learned various skills such as time management, how to collect clean data, process data, and set up a machine learning model. We also refined our programming skills by working with languages like HTML, CSS, JavaScript, and FastAPI.