search

jonfrank / football-predictor

AiCore project - football result prediction
2 stars 0 forks source link

How to update cleaned_dataset #1

Open owaio opened 2 years ago

owaio commented 2 years ago

Nice work BRO! I was tried use with your model, MLP, KNN and Ada, they are both above 60%. But now I am facing one question. How to update the cleaned_dataset.csv to latest matches. Maybe it could help improve the accuracy. Thanks for your mention.

jonfrank commented 2 years ago

Hey @owaio

Are you asking about creating additional rows in the training data for more recent matches so that the models could be re-trained, perhaps even weekly? I guess that's possible, but it wasn't the scope of the exercise. I think as well that, without a parameter to give more weight to more recent results, adding data week by week onto 30 years of historical data isn't going to make much difference in the short-term. (Mind you, that parameter to take recency into account is probably something that should be incorporated anyway, since arguably what happened between Arsenal and Man Utd in 1992 isn't particularly relevant to what will happen on Saturday.)

This was a project from my participation in AiCore - I can't tell whether you're on AiCore yourself too, or just stumbled across this repo randomly?!

owaio commented 2 years ago

Hey @jonfrank Bro,

Yup, I am asking for how to re-train the model when matches update by weekly. According your explain I believe it may not improve much more accuracy. That's my mind can help to improve data only. Btw, are there any method to add more league, like Asia(Japan, Korea, Australia) and America(Mexico, USA, Brazil and Chile)?

Sorry for my bad English.