MyVoterWisdom

2028 presidential county projections · trained on 2004–2024 historical data · 1,956 counties · 51 states

GitHub Open in Colab AIQualityManifesto

The Wisdom Formula

Completeness + Consistency + Validity = Wisdom
Completeness — Did we evaluate the entire context? Consistency — Are outcomes stable across 4 models? Validity — Are results grounded in historical data?

2028 County Prediction

Running 4 models — Random Forest, Logistic Regression, SVM, Gradient Boosting…
First request after idle may take up to 60s on the free tier.

Wisdom Analysis

Explore in Google Colab — free, no install required

✅ Dataset Quality Score

Validates prediction_pres_data.csv against the Wisdom Formula dimensions using the SysWisdom Data Quality API. The API key stays on the server — it is never exposed to your browser.

🗺️ 2028 Electoral College Projection

Runs the global Random Forest model over all 1,956 counties, aggregates projected county winners to state level, then maps to Electoral College votes. 270 electoral votes needed to win. Historical data only — not a forecast.

About this project

🏛️
Public data only

County-level election records 2004–2024. No PII. No voter targeting.

🔬
4 ML models

Random Forest · Logistic Regression · SVM · Gradient Boosting with SMOTE balancing

⚖️
Non-partisan

Educational research only. Not for use in campaigns or political advertising.

Data sources: 39 hand-curated counties (Tier 1, includes registration + mail-in data) combined with MEDSL / Harvard Dataverse presidential returns (Tier 2, CC BY 4.0) for 1,956 counties total. The original 39 counties were selected pragmatically for Wisdom Formula validation — not as a statistically representative sample. The global model now draws on cross-county patterns from all 1,956 counties. See full methodology →