About This System

How the election prediction model works

How It Works

Machine Learning Model

This system uses a Random Forest Classifier (300 decision trees) trained on 72 years of real Maharashtra electoral and assembly election data (1952–2024). Each tree votes on the outcome; the final win probability is calculated across all trees.

Recency Weighting

Not all elections are equal — politics changes over time! The model uses exponential decay weighting so that recent elections matter more:

2024 Election
1.00×
2022 Election
0.85×
2020 Election
0.72×
1952 Election
~0.00×

What the Model Looks At (10 Features)

  • Party Identity: Encoded party name — model learns each party's historical pattern
  • MLA Strength: How many MLAs the party currently holds
  • Alliance Strength: Combined MLA count including all coalition partners
  • Alliance Majority Flag: Whether the alliance holds ≥145 MLAs — the key majority threshold
  • MLA Share %: Party's share of all 288 Maharashtra seats (normalised)
  • Alliance Share %: Alliance's share of all 288 seats (normalised)
  • Past Seat Wins: Historical seat victory track record
  • Candidate Type: New, Experienced/Incumbent, or Mixed
  • Year: Captures political era trends (with recency weighting)

Model Stats

Algorithm
Random Forest
Decision Trees
300
Total Records
—
Features Used
10
Predicts
2027 Outcome

Disclaimer

Predictions are based on historical patterns only. Real elections can be affected by many factors not captured here — current events, campaign strategies, voter sentiment, etc. Use this as an informational tool, not a definitive forecast.