Κατεύθυνση / ειδίκευση Διαχείριση Πληροφορίας και Δεδομένων (ΔΕΔ)
Library of the School of Science

2016-07-29

2016

Γκάτση Ένρι
Mema Gerald

Ιωάννης Χαμόδρακας,Ιωάννης Eμίρης

Βελτίωση αλγορίθμων παραγωγής προτάσεων πολλαπλών κριτηρίων που βασίζονται στο συνεργατικό φιλτράρισμα

Greek

This thesis is an extension of G. Bourtsoukli’s thesis (2014),
which implemented a rating prediction algorithm to evaluate items according to
multiple
criteria, in order to recommend these items to the users according to their
preferences.
This thesis focuses on two aspects to improve the results of the algorithm.
First, a
modification of the method of linear regression which is used to calculate the
weights
assigned by the users to each criterion is proposed in order to improve its
results and
overcome its limitations. Furthermore, the problem of handling a larger number
of users
is faced by using an algorithm which finds approximate nearest neighbors by LSH
(locality sensitive hashing). LSH achieves the reduction of the dimensions of
the
problem and improves time and space complexity. Last but not least, the modified
algorithm is evaluated and compared with regard to its effectiveness and
performance
with two other popular algorithms, the Decomposing Multi Criteria and Weighted
Slope
One, as well as the initial algorithm

Personalization, Collaborative Filtering, Rating Estimation, Users Similarities, Evaluation Metrics

Yes

5-6

Yes

11

44

https://pergamos.lib.uoa.gr/uoa/dl/object/1324130/file.pdf

https://pergamos.lib.uoa.gr/uoa/dl/object/1324130