This is a semester-long project for my class in artificial intelligence.
In this program, I will be seeking to train a neural network that will be able to accurately predict whether or not a DotA 2 match was won or not. As a former dota 2 gamer, the ultimate test of this program will be to see if it can accurately predict the match results of my own previous games.
The datasets I am using to train this program can be found here on Kaggle I will also be using my own API key to retrieve my player's match history to test this program on. Feel free to do the same if you have an account with Valve.
Perhaps the most complicated aspect of this problem is first selecting which attributes to train this model on. I can draw upon my own experiences playing Dota to say that the team's gold and xp is important for winning, but this doesn't give the entire picture. It also introduces personal bias to the program's training. After an MVP is working, I would like to test different combinations of attributes to train the model on, regardless of my personal opinion. Perhaps Tower damage is the key to success here.
To be announced.
To be announced