Second Time posting this. ( First one is in general thread )
For those who don't know about recommender system or have only slight idea, read after the question explanation..
For ones that know what recommender system is... I saw around 38K+ titles on mangadex and an excellent tagging system to signify reading, completed, plan to read, dropped etc. which can significantly make the reduce the work required to make a recommender system. I know there are various categories in mangaupdates but those are simply based on the manga types and not on user preferences.
So questions:
1. Does this kind of system exits where i feed in what i have read based on bookmarks and stuff and get suggested mangas?
2. If not then how many users are there in mangadex? and is it possible to get that bookmark tag *data so that we can develop something on out own?
*The data will not require username or password , it will require only what different users have bookmarked keeping their username / account name anonymous ( say U1, has book this .. U2 has this )..
For everyone who is thinking what recommender system is... Recommender System is a way to suggest someone manga based on their similarity with other readers. That is how algorithm for netflix, amazon and youtube show some videos and movies which normally will be buried deep.
e.g: Say there are 4 persons A,B,C,D.
Manga Read by each one:-
A-->M1,M2,M3,M4,M5,M6,M7
B-->M1,M2,M3,M4,M5
C-->M1,M2,M3,M4,M5
D-->M1,M2,M3
Algorithm group them together so the suggested manga to user will be: ( The recommender system at work)
M6-->B,C
M7-->B,C
M4,M5-->D ( lets say D is relatively new to manga )
Depending on whether M6 or M7 is preferred by B and C, it will also be suggested to D later and thus the group of manga keeps on updating and helps the next user.
end of example.....
There are various simple algorithm, linear/nonlinear/logistic regression, random forest, svm among the simple ones to implement.. and neural network if we want it to be a bit more powerful.
It doesn't need to very high in accuracy since this will just be to help bring up some underrated Manga in suggestions, cause most top weekly are same and there is low chances of discovering some hidden gems.
So what does everyone say about recommender system?
For those who don't know about recommender system or have only slight idea, read after the question explanation..
For ones that know what recommender system is... I saw around 38K+ titles on mangadex and an excellent tagging system to signify reading, completed, plan to read, dropped etc. which can significantly make the reduce the work required to make a recommender system. I know there are various categories in mangaupdates but those are simply based on the manga types and not on user preferences.
So questions:
1. Does this kind of system exits where i feed in what i have read based on bookmarks and stuff and get suggested mangas?
2. If not then how many users are there in mangadex? and is it possible to get that bookmark tag *data so that we can develop something on out own?
*The data will not require username or password , it will require only what different users have bookmarked keeping their username / account name anonymous ( say U1, has book this .. U2 has this )..
For everyone who is thinking what recommender system is... Recommender System is a way to suggest someone manga based on their similarity with other readers. That is how algorithm for netflix, amazon and youtube show some videos and movies which normally will be buried deep.
e.g: Say there are 4 persons A,B,C,D.
Manga Read by each one:-
A-->M1,M2,M3,M4,M5,M6,M7
B-->M1,M2,M3,M4,M5
C-->M1,M2,M3,M4,M5
D-->M1,M2,M3
Algorithm group them together so the suggested manga to user will be: ( The recommender system at work)
M6-->B,C
M7-->B,C
M4,M5-->D ( lets say D is relatively new to manga )
Depending on whether M6 or M7 is preferred by B and C, it will also be suggested to D later and thus the group of manga keeps on updating and helps the next user.
end of example.....
There are various simple algorithm, linear/nonlinear/logistic regression, random forest, svm among the simple ones to implement.. and neural network if we want it to be a bit more powerful.
It doesn't need to very high in accuracy since this will just be to help bring up some underrated Manga in suggestions, cause most top weekly are same and there is low chances of discovering some hidden gems.
So what does everyone say about recommender system?
