OLLI WORKSHOP ON COMPUTERS VERSUS HUMANS

Second Sets of Visuals (September 21, 2010)
Copyright ©2010 by Theo Pavlidis


How Amazon matches books to readers (Collaborative Filtering)


Using Ratings from Several Users to Predict New Ratings

User Ratings for Two Items
User
Item 1
Item 2
John
4
5
Richard
4
4
Tiffany
2
3
Lucy
4
3
Michael
5
4
Basil
4
3
Betty
3
3
Data of the Table as a Graph
Regression Line

Using the Regression Line Piecewine Linear Regression Parabolic Regression

Opposite Ratings A Regression Curve and its use .... to measure error between prediction and reality.

Distance between Users
Distance(U1, U2)  =  |R1(A)-R2(A)| + |R1(B)-R2(B)| + ... + |R1(X)-R2(X)| + ....
Distance(John, Richard) 0 + 1 = 1 Distance(Tiffany, Lucy) 2 + 0 = 2
Distance(John, Tiffany) 2 + 2 = 4 Distance(Tiffany, Michael) 3 + 1 = 4
Distance(John, Lucy) 0 + 2 = 2 Distance(Tiffany, Basil) 2 + 0 = 2
Distance(John, Michael) 1 + 1 = 2 Distance(Tiffany, Betty) 1 + 0 = 1
Distance(John, Basil) 0 + 2 = 2 Distance(Lucy, Michael) 1 + 1 = 2
Distance(John, Betty) 1 + 2 = 3 Distance(Lucy, Basil) 0 + 0 = 0
Distance(Richard, Tiffany) 2 + 1 = 3 Distance(Lucy, Betty) 1 + 0 = 0
Distance(Richard, Lucy) 0 + 1 = 1 Distance(Michael, Basil) 1 + 1 = 2
Distance(Richard, Michael) 1+ 0 = 1 Distance(Michael, Betty) 2 + 1 = 3
Distance(Richard, Basil) 0 + 1 = 1 Distance(Basil, Betty) 1 + 0 = 1
Distance(Richard, Betty) 1 + 1 = 2