290 likes | 404 Views
Incorporating Reliability in a TV Recommender. Verus Pronk. Context. Increasing availability of TV programs Availability of electronic program guides (EPGs). How about a personal TV recommender? Applications Highlights in EPG Auto-recording/deletion on HD recorders
E N D
Incorporating Reliabilityin a TV Recommender Verus Pronk
Context • Increasing availability of TV programs • Availability of electronic program guides (EPGs) How about a personal TV recommender? Applications • Highlights in EPG • Auto-recording/deletion on HD recorders • Creation of personalized channels
Summary Introduction Naive Bayesian classification An example Reliable classification Results Concluding remarks
Introduction Thousands of programs offered each day People tend to browse only a limited number of channels EPGs provide easier access Low percentage of interesting programs More advanced solutions required
Introduction Programs are described by metadata (EPG) User rates a number of programs as J or L User profile describes relation between them
Introduction Example of metadata An Officer and a Gentleman: ( date : Tuesday, Nov. 23, 2004; time : 20:30 h.; station : SBS 6; genre : drama; cast : Richard Gere; credit : Taylor Hackford; ... )
Naive Bayesian classification Given : a training set X : i-th feature value of x known class of x Given : an instance t Asked : c(t) Approach: estimate based on the user profile calculated from X
Naive Bayesian classification Problem issues • Cold start • Changing preferences • Feature selection • Accuracy • Reliability • ...
posterior probabilities conditional probabilities prior probabilities Naive Bayesian classification
Naive Bayesian classification Conditional independence violation • The BBC news is always broadcast on the BBC • Clint Eastwood generally plays in action movies NBC is nevertheless successfully applied in many application areas
Naive Bayesian classification Priors set to pj Conditionals estimated using training set Denominator irrelevant
Naive Bayesian classification User profile
Naive Bayesian classification Classification error E is a convex combination of the Ejs
Naive Bayesian classification On the prior probabilities J
An example +1 +1 +1 +1 +1
Training set: 100 JTV programs 100 LTV programs Program:Tue. 20:30 Drama R. Gere T. Hackford : : J L
X X statistical analysis Reliable classification X random N(i, v, j) and N( j) random and dependent X uniform both binomially distributed
Reliable classification Theorem 1 Let Z ~ Bin(N, p), 0 < p < 1, Yn~Bin(n, q) Z0: Then ...
Reliable classification where
Reliable classification Theorem 2 Let Ri, i = 1, 2, ..., f, independent rconstant Then (Ris not actually independent)
Reliable classification Back to the original problem
Reliable classification Standard deviation of can be estimated by
Reliable classification Confidence intervals for Two approaches A: Fix k and don’t classify if intervals overlap: coverage B: Choose k such that intervals just do not overlap: explicit notion of confidence
Results Simulation TV recommender Training sets Briarcliff data Prior probabilities Set such that EJELE Confidence levelsk = 0, 0.1, 0.2,..., 1 Training set sizes100, 400 Approach A offset classification error against coverage
Concluding remarks • Reliability adds another dimension to classification • Our approach is explicit and robust • Separates difficult from easy instances • Also applicable to other domains • medical diagnosis • biometrics (e.g. face recognition) Acknowledgements Srinivas Gutta, Wim Verhaegh, Dee Denteneer