Credyt

Inspiration

Our parents have heard horror stories of people losing their funds to crypto scams and even falling victim during a few transactions. Despite the promise of trust and security, a lot of websites and transactions are unable to promise decentralization. We wanted to create a solution that preserves privacy and empowers users to control their financial decisions. We want to reduce their anxiety. One of our friends, who was in anger after losing a bet, accidentally sent $500 in bitcoin instead of the intended $50. Through Credyt, we hope to prevent problems like this occurring. We aim to build confidence and help people feel safe when interacting with the Web3 space. Ensuring everyone is in the right state of mind before any transaction. We hope Credyt helps individuals and businesses feel safe when using crypto.

What it does

Regression model uses factors and history from a user to generate and update their crypto credit score. In order to make this value more sensitive and dynamic, we included a nested model that calculates more parameters for the original model to use during inference. When a user navigates to the transaction page, a camera opens to detect the user's state of mind. It will prevent the user from making any transactions depending on its mood to ensure emotions don’t impact financial decisions.

How we built it

We outlined the project and divided it up into three technologically challenging aspects and each individually created elegant solutions to each of these facets on paper prior to implementing them. Following this, we integrated the functionality of our components and developed a modern frontend UI to parallel our advanced, innovative backend. We used chad-scn to get some of the components for our user interface. In addition to this, we used Flask and called APIs to connect our models to the frontend. For the backend, we made a regression model to calculate the crypto current score. Our classification model is what categorizes the transaction to detect any anomalies. By calling the APIs, we were able to output the result of the transaction for the user. We also

Challenges we ran into

One major technical challenge we faced was identifying a parameter set for our classification model to detect fraudulent transactions with minimal information. We had to overcome this by developing our own predictive heuristic to predict the purpose of a transaction. Another significant challenge we didn’t foresee was the lack of training data because we were creating a credit metric that didn't previously exist.

Accomplishments that we're proud of

One big accomplishment was recognizing the impact of emotions on financial decisions. This led us to add an advanced adding transaction feature that detects a users’ emotion to ensure if they’re in the right state of mind before making key financial decisions. We’re also proud of targeting a problem that has no existing solution. Most transactional

What we learned

Our biggest takeaway was learning about the importance of a trust framework for Web3. It's reputation cannot easily be defined on-chain and creating a reputable system doesn't just require cryptograhic methods but it also requires an understanding of BEHAVIOR and machine learning. That is why we used mutiple models while also detecting a person's mood to ensure they're in the right state of mind to make the right decisions.

What's next for Credyt

We want to use our credyt crypto score to influence people’s general credit scores, so that we can truly gauge how trustworthy a loaner is by including all of their financial behavior and history.

Built With

• blockchain
• coingecko
• deeplearning
• heuristics
• isolationforest
• keras
• nextjs
• opencv
• python
• regression
• tailwind
• tensorflow
• typescript
• xgboost
• zero-knowledge-proof