Random Number Pattern Detection Model
Budget: £10 – £15 GBP
I have several large spreadsheets holding mixed numerical-and-text strings that record the behaviour of a proprietary random-number generator over time. Your task is to dig into this historical data, uncover any repeatable structure, bias or correlation, and turn those insights into a robust pattern-detection model that I can run on fresh streams of numbers.
Scope
• Import and clean the spreadsheets, preserving every column exactly as logged.
• Explore the mixed data thoroughly—numerics, timestamps, metadata notes—and visualise anything that hints at non-random behaviour.
• Build and document a reproducible workflow (Python, R or another open-source stack) that flags statistically significant patterns and offers a clear metric of pattern strength.
• Provide an interactive notebook or script I can rerun with new data, plus a short report interpreting the results in plain language.
Acceptance criteria
1. All raw files load without manual fixes.
2. Statistical tests (e.g., chi-square, runs tests, spectral analysis) demonstrate whether detected patterns deviate from true randomness at the agreed significance level.
3. Code is commented, version-controlled and runnable on a standard desktop environment.
If you have experience reverse-engineering RNGs or working with high-entropy data, you’ll feel right at home here. Let me know the libraries you prefer and any similar projects you’ve cracked before.
Scope
• Import and clean the spreadsheets, preserving every column exactly as logged.
• Explore the mixed data thoroughly—numerics, timestamps, metadata notes—and visualise anything that hints at non-random behaviour.
• Build and document a reproducible workflow (Python, R or another open-source stack) that flags statistically significant patterns and offers a clear metric of pattern strength.
• Provide an interactive notebook or script I can rerun with new data, plus a short report interpreting the results in plain language.
Acceptance criteria
1. All raw files load without manual fixes.
2. Statistical tests (e.g., chi-square, runs tests, spectral analysis) demonstrate whether detected patterns deviate from true randomness at the agreed significance level.
3. Code is commented, version-controlled and runnable on a standard desktop environment.
If you have experience reverse-engineering RNGs or working with high-entropy data, you’ll feel right at home here. Let me know the libraries you prefer and any similar projects you’ve cracked before.