ESCPredictive

The Eurovision Predictive Model

Predicting Eurovision through data and machine learning.

📊 Full Split Results 🠚

🔄 Latest Update 🠚

What is ESCPredictive?

ESCPredictive is a machine learning project designed to predict Eurovision Song Contest results using historical data and statistical patterns.

By analysing past contests, the model identifies relationships between different factors and voting outcomes to estimate how countries might perform in future editions.

The goal is to combine data science with a passion for Eurovision to uncover insights that go beyond personal opinions and fan predictions.

From data to predictions

Data Collection

Historical Eurovision results and relevant contestant and country information are collected and prepared for analysis.

Feature Preprocessing

The available information is transformed into meaningful variables that help the model identify patterns associated with voting behaviour.

Nonlinear Regression

A regression model identifies broader statistical relationships between the expectations features and voting outcomes, providing one perspective on the expected results.

Gradient Boosting

A gradient boosting model captures more complex, nonlinear relationships and interactions within the musical and artistic features, offering a complementary perspective.

Hybrid Ensemble

The outputs of both models are combined into an ensemble prediction, bringing together their complementary strengths.

Points Allocation

The ensemble predictions are integrated with the voting database to simulate the points awarded to each country and produce a predicted scoreboard.

Two Models. One Prediction.

Linear Regression + XGBoost = A Hybrid Ensemble

Why combine two models instead of relying on just one?

ESCPredictive combines two complementary machine learning models, each designed to analyse a different side of Eurovision. Nonlinear Regression uses expectation-based features, such as betting odds and chart performance, to capture existing predictions of how each entry might perform. XGBoost, on the other hand, analyses artistic and musical features to identify patterns that may influence voting outcomes.

By combining their predictions, ESCPredictive brings together market expectations and the characteristics of each entry to produce a more comprehensive estimate of Eurovision voting outcomes.

The resulting hybrid ensemble model goes beyond predicting rankings. Its outputs are fed into a historical voting database that replicates Eurovision's voting system, using the ensemble predictions alongside historical voting patterns to calculate the points awarded to each country by the public and national juries. This process ultimately produces a complete predicted Eurovision scoreboard.