TrendScore

• 

  Features and Outputs

Inspiration

We kept noticing the same gap: creators can see what is trending, but they still do not know if it is actually worth their time. We wanted to build something that helps people make that call before they spend hours making a video.

What it does

TrendScore helps YouTube creators and agencies decide whether a trend is worth jumping on. It gives a trend worthiness score, timing guidance, and a simple recommendation based on pre-publish signals.

How we built it

We used a US YouTube Trending dataset, cleaned it, engineered useful features, and built a two-stage ML pipeline. K-Means helped us group trends into archetypes, and HistGradientBoostingClassifier helped us score whether a trend is worth joining. Then we turned it into a working app using FastAPI, Streamlit, Plotly, React, and TypeScript.

Challenges we ran into

One of the hardest parts was defining what “worth joining” really means in a way a model can learn. We also had to deal with noisy engagement data and make sure the final product stayed simple, explainable, and useful.

Accomplishments that we're proud of

We are proud that TrendScore is more than just a model. We built a working end-to-end product with a live demo, solid model performance, clear trend archetypes, and a real interface people can use.

What we learned

We learned that accuracy alone is not enough. A good ML project also needs to be understandable, practical, and trustworthy. We also learned a lot about cleaning messy data, choosing the right features, and turning technical results into something useful for real users.

What's next for TrendScore

Next, we want to improve the scoring logic, expand beyond the US dataset, and make the recommendations more personalized. We also want to support more creator workflows and keep making the product more useful in real publishing decisions.

Built With

• fastapi
• histgradientboostingclassifier
• including-k-means-and-histgradientboostingclassifier-fastapi-and-uvicorn-for-the-backend-api-streamlit-and-plotly-for-the-interactive-dashboard-react
• k-means
• plotly
• python
• react
• render
• scikit-learn
• streamlit
• typescript
• uvicorn
• vite