Inspiration
Environmental sensor datasets can contain missing measurements, repeated timestamps, implausible readings and unexplained spikes. Decisions based on these data should start with transparent quality checks.
What it does
ClimateProof is a Python/Streamlit prototype that imports timestamped CSV observations, highlights missing or nonnumeric values, missing timestamps, duplicate timestamps, optional user-defined plausibility violations and robust statistical outliers. It displays a time-series chart and exports a flagged CSV plus a JSON quality report. It calculates a SHA-256 fingerprint of the original uploaded file bytes.
How we built it
Streamlit provides the UI; pandas handles CSV parsing, time series and data checks; Python hashlib computes the SHA-256 digest. Modified z-scores based on median absolute deviation flag exploratory outliers, with zero-MAD cases handled conservatively.
Challenges and lessons
Data integrity is not data authenticity. A digest only detects a file change when compared with a previously trusted fingerprint. Statistical anomalies do not automatically mean faulty sensors. Bounds must be selected based on the sensor and unit.
Accomplishments
A runnable local prototype, sample dataset, setup instructions and downloadable analysis reports have been prepared. The demo, public GitHub repository and final submission are not yet complete.
What's next
Publish the repository, run a recorded 3–5-minute demonstration and submit the entry. Future versions could support sensor metadata, expected sampling intervals, reference datasets and trusted provenance workflows.
Limitations
This is a prototype tested syntactically, not a production monitoring system. No blockchain anchoring, real-world deployment or verified climate impact is claimed.
Log in or sign up for Devpost to join the conversation.