Personalized Channel Recommendation Deep Learning from a Switch Sequence

Can Yang, Sixuan Ren, Yong Liu, Houwei Cao, Qihu Yuan, Guoqiang Han

Research output: Contribution to journalArticle

Abstract

Internet protocol TV (IPTV) services could enhance personalized viewing experience in a more interactive way than traditional broadcast TV systems, but it is still difficult for subscribers to quickly find interesting channels to watch from a huge selection. This paper focuses on a framework for personalized live channel recommending via deep learning from a historical switching sequence with a long short-term memory (LSTM) neural network. Using real-world IPTV watching logs, we first obtained insights into user behaviors when watching live channels, and then proposed a learning scheme on how to dynamically generate a recommended channel list for each user with an independent LSTM net trained using the channel watching history during a slide window. For designing a good data architecture and representation scheme for a dynamically learning framework, we then studied the performance of the proposed recommendation method by varying the width of the slide window for training, the length of input sequence for prediction, and the mode to process input and label space. We finally developed a separate learning method to fairly recommend for popular (hot) or unpopular (cold) channels, respectively, based on channel popularity in the training set with an extra price of a possible hit lag after recommendation, in order to alleviate the Matthew effect arising from the conventional recommendation based on historical information. The experimental results show LSTM succeeds in learning from a historical channel switching sequence, outperforms several baseline recommendation methods, especially for hot channels, and the classified recommendation by separate learning brings an overall performance gain.

Original languageEnglish (US)
Article number8458124
Pages (from-to)50824-50838
Number of pages15
JournalIEEE Access
Volume6
DOIs
StatePublished - Sep 7 2018

Fingerprint

Internet protocols
Switches
Labels
Neural networks
Long short-term memory
Deep learning

Keywords

  • Deep learning
  • IPTV
  • long-short term memory
  • recommender systems
  • recurrent neural networks
  • separate learning
  • user behavior analysis

ASJC Scopus subject areas

  • Computer Science(all)
  • Materials Science(all)
  • Engineering(all)

Cite this

Personalized Channel Recommendation Deep Learning from a Switch Sequence. / Yang, Can; Ren, Sixuan; Liu, Yong; Cao, Houwei; Yuan, Qihu; Han, Guoqiang.

In: IEEE Access, Vol. 6, 8458124, 07.09.2018, p. 50824-50838.

Research output: Contribution to journalArticle

Yang, Can ; Ren, Sixuan ; Liu, Yong ; Cao, Houwei ; Yuan, Qihu ; Han, Guoqiang. / Personalized Channel Recommendation Deep Learning from a Switch Sequence. In: IEEE Access. 2018 ; Vol. 6. pp. 50824-50838.
@article{2bbfd8822c554810aef54010c00aed95,
title = "Personalized Channel Recommendation Deep Learning from a Switch Sequence",
abstract = "Internet protocol TV (IPTV) services could enhance personalized viewing experience in a more interactive way than traditional broadcast TV systems, but it is still difficult for subscribers to quickly find interesting channels to watch from a huge selection. This paper focuses on a framework for personalized live channel recommending via deep learning from a historical switching sequence with a long short-term memory (LSTM) neural network. Using real-world IPTV watching logs, we first obtained insights into user behaviors when watching live channels, and then proposed a learning scheme on how to dynamically generate a recommended channel list for each user with an independent LSTM net trained using the channel watching history during a slide window. For designing a good data architecture and representation scheme for a dynamically learning framework, we then studied the performance of the proposed recommendation method by varying the width of the slide window for training, the length of input sequence for prediction, and the mode to process input and label space. We finally developed a separate learning method to fairly recommend for popular (hot) or unpopular (cold) channels, respectively, based on channel popularity in the training set with an extra price of a possible hit lag after recommendation, in order to alleviate the Matthew effect arising from the conventional recommendation based on historical information. The experimental results show LSTM succeeds in learning from a historical channel switching sequence, outperforms several baseline recommendation methods, especially for hot channels, and the classified recommendation by separate learning brings an overall performance gain.",
keywords = "Deep learning, IPTV, long-short term memory, recommender systems, recurrent neural networks, separate learning, user behavior analysis",
author = "Can Yang and Sixuan Ren and Yong Liu and Houwei Cao and Qihu Yuan and Guoqiang Han",
year = "2018",
month = "9",
day = "7",
doi = "10.1109/ACCESS.2018.2869470",
language = "English (US)",
volume = "6",
pages = "50824--50838",
journal = "IEEE Access",
issn = "2169-3536",
publisher = "Institute of Electrical and Electronics Engineers Inc.",

}

TY - JOUR

T1 - Personalized Channel Recommendation Deep Learning from a Switch Sequence

AU - Yang, Can

AU - Ren, Sixuan

AU - Liu, Yong

AU - Cao, Houwei

AU - Yuan, Qihu

AU - Han, Guoqiang

PY - 2018/9/7

Y1 - 2018/9/7

N2 - Internet protocol TV (IPTV) services could enhance personalized viewing experience in a more interactive way than traditional broadcast TV systems, but it is still difficult for subscribers to quickly find interesting channels to watch from a huge selection. This paper focuses on a framework for personalized live channel recommending via deep learning from a historical switching sequence with a long short-term memory (LSTM) neural network. Using real-world IPTV watching logs, we first obtained insights into user behaviors when watching live channels, and then proposed a learning scheme on how to dynamically generate a recommended channel list for each user with an independent LSTM net trained using the channel watching history during a slide window. For designing a good data architecture and representation scheme for a dynamically learning framework, we then studied the performance of the proposed recommendation method by varying the width of the slide window for training, the length of input sequence for prediction, and the mode to process input and label space. We finally developed a separate learning method to fairly recommend for popular (hot) or unpopular (cold) channels, respectively, based on channel popularity in the training set with an extra price of a possible hit lag after recommendation, in order to alleviate the Matthew effect arising from the conventional recommendation based on historical information. The experimental results show LSTM succeeds in learning from a historical channel switching sequence, outperforms several baseline recommendation methods, especially for hot channels, and the classified recommendation by separate learning brings an overall performance gain.

AB - Internet protocol TV (IPTV) services could enhance personalized viewing experience in a more interactive way than traditional broadcast TV systems, but it is still difficult for subscribers to quickly find interesting channels to watch from a huge selection. This paper focuses on a framework for personalized live channel recommending via deep learning from a historical switching sequence with a long short-term memory (LSTM) neural network. Using real-world IPTV watching logs, we first obtained insights into user behaviors when watching live channels, and then proposed a learning scheme on how to dynamically generate a recommended channel list for each user with an independent LSTM net trained using the channel watching history during a slide window. For designing a good data architecture and representation scheme for a dynamically learning framework, we then studied the performance of the proposed recommendation method by varying the width of the slide window for training, the length of input sequence for prediction, and the mode to process input and label space. We finally developed a separate learning method to fairly recommend for popular (hot) or unpopular (cold) channels, respectively, based on channel popularity in the training set with an extra price of a possible hit lag after recommendation, in order to alleviate the Matthew effect arising from the conventional recommendation based on historical information. The experimental results show LSTM succeeds in learning from a historical channel switching sequence, outperforms several baseline recommendation methods, especially for hot channels, and the classified recommendation by separate learning brings an overall performance gain.

KW - Deep learning

KW - IPTV

KW - long-short term memory

KW - recommender systems

KW - recurrent neural networks

KW - separate learning

KW - user behavior analysis

UR - http://www.scopus.com/inward/record.url?scp=85053126007&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85053126007&partnerID=8YFLogxK

U2 - 10.1109/ACCESS.2018.2869470

DO - 10.1109/ACCESS.2018.2869470

M3 - Article

VL - 6

SP - 50824

EP - 50838

JO - IEEE Access

JF - IEEE Access

SN - 2169-3536

M1 - 8458124

ER -