NEON VEIL

Inspiration:

Halloween meets data science. I wanted to turn ordinary anomaly detection into something immersive—a “ghost-hunting lab” where each spike, whisper, or temperature drop tells a story. NEON VEIL was born from the idea that learning AI can be cinematic, poetic, and spooky all at once.

What it does:

NEON VEIL simulates a haunted-house monitoring system that tracks four sensors—EMF, temperature, sound, and motion—over an 8-hour night.

It then detects strange “paranormal” patterns using: Rolling z-scores for quick statistical anomalies. Isolation Forest for unsupervised ML detection.

The notebook outputs: Dynamic plots of ghostly spikes and cold spots. CSVs of sensor logs and detected events. An auto-generated Investigation Report summarizing the night’s findings.

How we built it:

Language & Tools: Python · NumPy · Pandas · Matplotlib · SciPy · scikit-learn Environment: Jupyter Notebook (offline, fully reproducible) Design: Neon-purple & dark-themed visuals to fit the “spectral lab” aesthetic. Automation: Generates Markdown reports and CSV exports automatically.

Challenges we ran into:

Tuning thresholds to make “ghost events” feel believable without overwhelming noise. Balancing the spooky narrative with explainable ML logic. Making the notebook visual yet lightweight for any machine to run.

Accomplishments that we're proud of:

A creative, educational ML demo that runs 100 % offline. Cinematic data visualization of AI-detected “hauntings.” A Halloween project that doubles as a fun tutorial on time-series anomaly detection.

What we learned:

How storytelling can make technical concepts stick. Practical ML for unsupervised anomaly detection. How small aesthetic touches—color, rhythm, narrative—can transform data into experience.

What's next for NEON VEIL:

Add a real-time stream mode that triggers “presence detected” alerts. Build a Streamlit UI with glowing neon controls. Extend the ML layer with clustering and cross-sensor correlation. Host a “Ghost Dataset” leaderboard so others can test anomaly models on spooky data.

Built With

• jupyter
• matplotlibscipy
• numpy
• pandas
• python
• scikit-learn