Performance comparison of top N recommendation algorithms

Ghulam Mustafa; Ingo Frommholz

doi:10.1109/FGCT.2015.7300256

Performance comparison of top N recommendation algorithms

Ghulam Mustafa
, Ingo Frommholz

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

2 Citations (Scopus)

Abstract

In traditional recommender systems, services/items are recommended to the user based on the initial ratings while the results comes from the predicted rating values are not considered which further refers to top N recommendations. In top N recommendation algorithms, recommendation process is further enhanced by predicting the missing ratings where the basic objective is to find the items that might be interest of a user. Performance comparison and evaluation of different top N recommendation algorithms is quite challenging for large datasets where selection of an appropriate algorithm can help to improve the recommendation process by predicting missing ratings. Therefore, in this paper we analyse and evaluate the 6 different top N recommendation algorithms using accuracy metrics such as precision and recall on Movie-lense 100K dataset from the Group-lens. Our main finding is the selection of Top N recommendation algorithm that perform significantly better than other recommender algorithms in pursuing the top-N recommendation process.

Original language	English
Title of host publication	nan
Publisher	Institute of Electrical and Electronics Engineers Inc.
ISBN (Print)	9781479982660
DOIs	https://doi.org/10.1109/FGCT.2015.7300256
Publication status	Published - 26 Oct 2015
Event	2015 Fourth International Conference on Future Generation Communication Technology (FGCT) - Luton Duration: 29 Jul 2015 → 31 Jul 2015

Conference

Conference	2015 Fourth International Conference on Future Generation Communication Technology (FGCT)
City	Luton
Period	29/07/15 → 31/07/15
Other	2015 Fourth International Conference on Future Generation Communication Technology (FGCT) (29/07/2015-31/07/2015, Luton)

Keywords

Collaborative Filtering
Memory Based CF
Model Based CF
Principle Component Analysis (PCA) and Singular Value Decomposition (SVD)
ROC Curve
Recommender System

Access to Document

10.1109/FGCT.2015.7300256

https://ieeexplore.ieee.org/document/7300256

Handle.net

Persistent link

Cite this

@inproceedings{e00aa5cf19d94329a127f43c17440897,

title = "Performance comparison of top N recommendation algorithms",

abstract = "In traditional recommender systems, services/items are recommended to the user based on the initial ratings while the results comes from the predicted rating values are not considered which further refers to top N recommendations. In top N recommendation algorithms, recommendation process is further enhanced by predicting the missing ratings where the basic objective is to find the items that might be interest of a user. Performance comparison and evaluation of different top N recommendation algorithms is quite challenging for large datasets where selection of an appropriate algorithm can help to improve the recommendation process by predicting missing ratings. Therefore, in this paper we analyse and evaluate the 6 different top N recommendation algorithms using accuracy metrics such as precision and recall on Movie-lense 100K dataset from the Group-lens. Our main finding is the selection of Top N recommendation algorithm that perform significantly better than other recommender algorithms in pursuing the top-N recommendation process.",

keywords = "Collaborative Filtering, Memory Based CF, Model Based CF, Principle Component Analysis (PCA) and Singular Value Decomposition (SVD), ROC Curve, Recommender System",

author = "Ghulam Mustafa and Ingo Frommholz",

year = "2015",

month = oct,

day = "26",

doi = "10.1109/FGCT.2015.7300256",

language = "English",

isbn = "9781479982660",

booktitle = "nan",

publisher = "Institute of Electrical and Electronics Engineers Inc.",

address = "United States",

note = "2015 Fourth International Conference on Future Generation Communication Technology (FGCT) ; Conference date: 29-07-2015 Through 31-07-2015",

}

TY - GEN

T1 - Performance comparison of top N recommendation algorithms

AU - Mustafa, Ghulam

AU - Frommholz, Ingo

PY - 2015/10/26

Y1 - 2015/10/26

N2 - In traditional recommender systems, services/items are recommended to the user based on the initial ratings while the results comes from the predicted rating values are not considered which further refers to top N recommendations. In top N recommendation algorithms, recommendation process is further enhanced by predicting the missing ratings where the basic objective is to find the items that might be interest of a user. Performance comparison and evaluation of different top N recommendation algorithms is quite challenging for large datasets where selection of an appropriate algorithm can help to improve the recommendation process by predicting missing ratings. Therefore, in this paper we analyse and evaluate the 6 different top N recommendation algorithms using accuracy metrics such as precision and recall on Movie-lense 100K dataset from the Group-lens. Our main finding is the selection of Top N recommendation algorithm that perform significantly better than other recommender algorithms in pursuing the top-N recommendation process.

AB - In traditional recommender systems, services/items are recommended to the user based on the initial ratings while the results comes from the predicted rating values are not considered which further refers to top N recommendations. In top N recommendation algorithms, recommendation process is further enhanced by predicting the missing ratings where the basic objective is to find the items that might be interest of a user. Performance comparison and evaluation of different top N recommendation algorithms is quite challenging for large datasets where selection of an appropriate algorithm can help to improve the recommendation process by predicting missing ratings. Therefore, in this paper we analyse and evaluate the 6 different top N recommendation algorithms using accuracy metrics such as precision and recall on Movie-lense 100K dataset from the Group-lens. Our main finding is the selection of Top N recommendation algorithm that perform significantly better than other recommender algorithms in pursuing the top-N recommendation process.

KW - Collaborative Filtering

KW - Memory Based CF

KW - Model Based CF

KW - Principle Component Analysis (PCA) and Singular Value Decomposition (SVD)

KW - ROC Curve

KW - Recommender System

U2 - 10.1109/FGCT.2015.7300256

DO - 10.1109/FGCT.2015.7300256

M3 - Conference contribution

SN - 9781479982660

BT - nan

PB - Institute of Electrical and Electronics Engineers Inc.

T2 - 2015 Fourth International Conference on Future Generation Communication Technology (FGCT)

Y2 - 29 July 2015 through 31 July 2015

ER -

Performance comparison of top N recommendation algorithms

Abstract

Conference

Keywords

Access to Document

Handle.net

Fingerprint

Cite this