search

kochlisGit / ProphitBet-Soccer-Bets-Predictor

ProphitBet is a Machine Learning Soccer Bet prediction application. It analyzes the form of teams, computes match statistics and predicts the outcomes of a match using Advanced Machine Learning (ML) methods. The supported algorithms in this application are Neural Networks, Random Forests & Ensembl Models.
MIT License
281 stars 104 forks source link

Differences between Evaluation Samples: Evaluation and Training mode #61

Open keith0811 opened 5 months ago

keith0811 commented 5 months ago

What is the difference between evaluation and training option under evaluation sample?

kochlisGit commented 5 months ago

Evaluation option is to "simulate" how well the model would perform in a scenario where you would use the model. Then, you can play with the filters and percentiles to see in which matches your model performs well (e.g It predicts with very high accuracy Home Wins when the odd 1 is betwen 1.30 and 1.60).

On the other hand, use the training option to check how well your model "fits" (learns) the training dataset. For instance, if the accuracy is too small, it means it has learnt correctly (Under-fitting). If the accuracy is too high, but the evaluation accuracy is small, it means that the model has learnt the training dataset so well that it fails to predict the outcomes of new unseen matches, which are similar to the training dataset (Over-fitting),

For example, if You train a Neural Network for 100 epochs and 5 hidden layers of 300-500 units each layer, you might notice a 100% percent accuracy in training set but only 30% on evaluation set, which means the model has overfitted.