The Hype Dashboard

AI-powered trend analyzer that predicts if a viral keyword is a lasting Fundamental or a passing Fad — using Bass Diffusion, Dynamic Time Warping, and ML on live Google Trends data in real time.

Comment

Inspiration

What it does

How we built it

Challenges we ran into

Accomplishments that we're proud of

What we learned

What's next for The Hype Dashboard## Inspiration

Every year, billions of dollars are lost chasing trends that turn out to be fads — from NFTs to Clubhouse to fidget spinners. We asked: can math and ML predict which trends have staying power before the crash?

What it does

The Hype Dashboard takes any keyword and runs a real-time analysis pipeline:

  1. Pulls live Google Trends search volume data
  2. Scores public sentiment using VADER NLP
  3. Checks financial trading signals for institutional backing
  4. Fits the data to the Bass Diffusion Model to measure innovation vs. imitation
  5. Compares the curve shape to 60 known fads & fundamentals using Dynamic Time Warping
  6. Classifies with a Random Forest trained on historical trend data

The output: a Fad Score (0–100), an estimated half-life, a prediction curve, and a full metrics breakdown — all on a sleek, animated dashboard.

How we built it

  • Backend: Python + FastAPI. The ML engine uses scipy for Bass Diffusion curve fitting, scikit-learn for Random Forest classification, and dtaidistance for DTW.
  • Data: Live Google Trends via pytrends, VADER sentiment analysis, synthetic financial correlation signals.
  • Frontend: Vanilla HTML/CSS/JS with a dark glassmorphism theme, animated SVG gauges, and Plotly.js interactive charts.
  • Baseline Database: 60 pre-computed trend curves (30 fads + 30 fundamentals) used to train the classifier.

Challenges we ran into

  • Bass Diffusion curve fitting is numerically unstable on noisy real-world data — we solved this by fitting on cumulative values instead of raw time series.
  • The q/p ratio can explode to absurd values when p approaches zero — we added display capping and heuristic fallbacks.
  • Google Trends API has aggressive rate limits — we built graceful fallback to synthetic data so the demo always works.

What we learned

  • The Bass Diffusion Model's q/p ratio is a surprisingly powerful single-feature predictor of fad vs. fundamental.
  • Dynamic Time Warping is excellent for comparing time-series shapes regardless of speed or timing differences.
  • Real-world trend data is messy — robust fallback strategies are essential for a reliable demo.

What's next

  • Integrate live Reddit (PRAW) and News API for real-time sentiment instead of synthetic contexts.
  • Add Alpha Vantage for live financial data correlation.
  • Deploy as a public web service with user accounts and trend tracking over time.

Built With

Share this project:

Updates