2
$\begingroup$

I have a project which is to predict in next time, which option player will select for betting. This prediction based on his history. For example, the player have time series data, the last options they selected, how much money they spend, did they loose or win...ect. There are many options for one bet, around 300. Based on that data, I want to predict next time which option (or possibility of options) they will choose for betting.

I don't know which kind of problem it is. Is it kind of product recommendation, multi-class classification or multi-label classification? Any keyword, or suggestion to solve this kind of problem? Thanks.

$\endgroup$
1
$\begingroup$

this is a multi class classification problem.

There are many approaches you can use to solve this problem.

You can first try Gradient Boosted Trees with libraries like xgboost or lightgbm. Both of those can be used in R or Python.

You can also use neural networks for multi class classification problems.

But there are many ways to solve a multi classification problem.

You can turn it into a multi binary classification problem. Since you have 300 options, you can train 300 binary classification algorithms on your data.

You can also try K nearest neighbours.

| improve this answer | |
$\endgroup$
  • $\begingroup$ is the 300 classes too much for K nearest neighbours or any linear model? Should I go with neural networks if the features are around 10, dataset is around 10-20 millions rows? $\endgroup$ – Quan Duong Feb 26 '18 at 14:38
  • $\begingroup$ I would definitely go with a neural network, 10 features is ok, but with amount of rows you have to make sure you have enough computing power. $\endgroup$ – user2505650 Feb 26 '18 at 15:28
1
$\begingroup$

Assuming you have betting history for multiple players, this is a collaborative filtering problem. I’d treat each bet as a separate transaction that has the characteristics price, class, and outcome. You can then predict the next class.

| improve this answer | |
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.