Download as pdf or txt
Download as pdf or txt
You are on page 1of 10

Effective Missing Data Prediction

Effective Missing Data Prediction

Bidyut Kr. Patra

National Institute of Technology Rourkela, Rourkela, India.

1/8
Effective Missing Data Prediction
Outline

2/8
Effective Missing Data Prediction
Outline

Effective Missing Data Prediction for CF (2007)


PCC based CF generally achieves higher performance than
other CF if dataset is dense.
PCC overestimates the similarity between two uses who
happen to have rated few items identically.

Herlocker proposed to used following formula:

max |Iu I u |,
sim(u, u ) = simPCC (u, u )

3/8
Effective Missing Data Prediction
Outline

Effective Missing Data Prediction for CF (2007)


PCC based CF generally achieves higher performance than
other CF if dataset is dense.
PCC overestimates the similarity between two uses who
happen to have rated few items identically.

Herlocker proposed to used following formula:

max |Iu I u |,
sim(u, u ) = simPCC (u, u )

min |Iu I u |,
sim(u, u ) = simPCC (u, u )

3/8
Effective Missing Data Prediction
Outline

Effective Missing Data Prediction for CF (2007)


PCC based CF generally achieves higher performance than
other CF if dataset is dense.
PCC overestimates the similarity between two uses who
happen to have rated few items identically.

Herlocker proposed to used following formula:

max |Iu I u |,
sim(u, u ) = simPCC (u, u )

min |Iu I u |,
sim(u, u ) = simPCC (u, u )

min |Ui Uj |,
sim(i , j) = simCorrelation (i , j)

3/8
Effective Missing Data Prediction
Outline

Effective Missing Data Prediction for CF (Hao Ma, 2007)

For each missing data rui , generate a set of similar users S(u) :
S(u) = {v U |sim(u, v ) > }
For each missing data rui , a set of simialr items S(i ):
S(i ) = {i I|sim(i , j) > }

4/8
Effective Missing Data Prediction
Outline

Effective Missing Data Prediction for CF (Hao Ma, 2007)

For each missing data rui , if S(u) 6= , S(i ) 6= ,


rui =P P
Sim(u,ua ).(rua i ua ) sim(i ,j).(ruj ij )
P
ua S(u) Sim(u,ua )+ (1 ) i+ jS(i
P)
sim(i ,j)
ua S(u) jS(i )

5/8
Effective Missing Data Prediction
Outline

Effective Missing Data Prediction for CF (Hao Ma, 2007)

For each missing data rui , if S(u) 6= , S(i ) 6= ,


rui =P P
Sim(u,ua ).(rua i ua ) sim(i ,j).(ruj ij )
P
ua S(u) Sim(u,ua )+ (1 ) i+ jS(i
P)
sim(i ,j)
ua S(u) jS(i )

S(u) 6=
P, S(i ) =
Sim(u,ua ).(rua i ua )
rui = P
ua S(u)
Sim(u,ua )
ua S(u)

S(u) = ,P S(i ) 6=
sim(i ,j).(ruj ij )
rui = i + jS(i
P)
sim(i ,j)
jS(i )

5/8
Effective Missing Data Prediction
Outline

Effective Missing Data Prediction for CF

S(i ) = , S(u) =

rui = ru + (1 ) ri

6/8
Effective Missing Data Prediction
Outline

THANK YOU

7/8

You might also like