data source
extreme4all opened this issue · 7 comments
hey, can i help making the flask api & NN?
Sure! Just submit post requests and I'll accept them if they're able to help contribute to the system! let me update the repo before you do that - I'll let you know once that's done!
great, yeah access to the data on your server and a format must be agreed upon
Oh yeah no worries! Nothing at all like that - I just need to make sure the files are all up to date so you can access them properly. Apologies btw, I'm not a programmer, so the code might look messy. I'm just gathering the last remaining files atm!
https://github.com/Ferrariic/Python-Bot-Detector
Here you go, these are all the files I think you would need. It includes everything that's being worked on atm.
Right now the method is:
Pull names from in game --> Look up on hiscores --> normalize --> run through Kmeans (n = 300 clusters for this first attempt) --> run this through the KNN --> Upload the KNN to the flask server and use that to classify new names that are input into the system.
I'm pretty much a novice at best when it comes to data science/classifiers/etc. So any help is greatly appreciated.
probably most work is spend on feature engineering :)
like if we have a flow to label banned users, aka users that we found on the highscores but later we cannot find them on the highscores (or i believe banned users are banned from the highscores)
Yeah that would be a great addition! What do you think about sampling users within a group, and then having automated weekly checks on those users to see if they still appear on the hiscores? We might be able to use that to evaluate ban rate - as one idea.