Research on Twitch Platform Channel Recommendation Based on User Usage Mining
碩士 === 國立臺中科技大學 === 資訊工程系碩士班 === 107 === In the era of online video and audio, more and more people are watching broadcasts which thereby creates great business opportunities for the broadcasters. Viewer clicks to watch are considered important behaviors to the broadcast. Therefore, this paper propo...
Main Authors: | , |
---|---|
Other Authors: | |
Format: | Others |
Language: | zh-TW |
Published: |
2019
|
Online Access: | http://ndltd.ncl.edu.tw/handle/24a6v4 |
id |
ndltd-TW-107NTTI5392021 |
---|---|
record_format |
oai_dc |
spelling |
ndltd-TW-107NTTI53920212019-09-24T03:34:25Z http://ndltd.ncl.edu.tw/handle/24a6v4 Research on Twitch Platform Channel Recommendation Based on User Usage Mining 植基於觀眾行為分析之Twitch平台頻道推薦系統 Chih-Yu Chen 陳熾宇 碩士 國立臺中科技大學 資訊工程系碩士班 107 In the era of online video and audio, more and more people are watching broadcasts which thereby creates great business opportunities for the broadcasters. Viewer clicks to watch are considered important behaviors to the broadcast. Therefore, this paper proposes a VBROC recommendation system that improves the click rate. The VBROC recommendation system is a recommendation system that is built by the viewer viewing behavior and the rank order clustering algorithm. The viewer viewing behavior is the broadcast and time viewed by the viewer where the number of times the viewer views a broadcast and percentages of the viewer''s views for each broadcaster were calculated. The percentage viewing behavior is called adhesive capacity. The rank order clustering algorithm takes the binary of the viewer views for a broadcast and then sorts the number of viewers watching the broadcast. The viewers will be sorted to the higher rank by the most watched broadcast and according to the number of viewers. After sorting is completed, it will be combined with the viewer viewing behavior, and the binary values in the rank order clustering algorithm will be replaced by the calculated stickiness. Then, through the scoring method proposed in this paper, the viewer''s adhesion is recalculated, and the broadcaster with high adhesion score becomes the recommended to view list. The viewer''s click rate is simulated through the viewer''s viewing record. The experiment was simulated on the Twitch platform for three-month viewing records and for a total of about 650 million viewers. Experimental results showed that the VBROC recommendation system has a significantly improved click-through rate compared to multiple algorithms. In the future, this paper hopes to provide a reference for planners of the broadcast programs as a base to order recommended broadcasts. 陳民枝 黃馨逸 2019 學位論文 ; thesis 55 zh-TW |
collection |
NDLTD |
language |
zh-TW |
format |
Others
|
sources |
NDLTD |
description |
碩士 === 國立臺中科技大學 === 資訊工程系碩士班 === 107 === In the era of online video and audio, more and more people are watching broadcasts which thereby creates great business opportunities for the broadcasters. Viewer clicks to watch are considered important behaviors to the broadcast. Therefore, this paper proposes a VBROC recommendation system that improves the click rate. The VBROC recommendation system is a recommendation system that is built by the viewer viewing behavior and the rank order clustering algorithm. The viewer viewing behavior is the broadcast and time viewed by the viewer where the number of times the viewer views a broadcast and percentages of the viewer''s views for each broadcaster were calculated. The percentage viewing behavior is called adhesive capacity. The rank order clustering algorithm takes the binary of the viewer views for a broadcast and then sorts the number of viewers watching the broadcast. The viewers will be sorted to the higher rank by the most watched broadcast and according to the number of viewers. After sorting is completed, it will be combined with the viewer viewing behavior, and the binary values in the rank order clustering algorithm will be replaced by the calculated stickiness. Then, through the scoring method proposed in this paper, the viewer''s adhesion is recalculated, and the broadcaster with high adhesion score becomes the recommended to view list. The viewer''s click rate is simulated through the viewer''s viewing record. The experiment was simulated on the Twitch platform for three-month viewing records and for a total of about 650 million viewers. Experimental results showed that the VBROC recommendation system has a significantly improved click-through rate compared to multiple algorithms. In the future, this paper hopes to provide a reference for planners of the broadcast programs as a base to order recommended broadcasts.
|
author2 |
陳民枝 |
author_facet |
陳民枝 Chih-Yu Chen 陳熾宇 |
author |
Chih-Yu Chen 陳熾宇 |
spellingShingle |
Chih-Yu Chen 陳熾宇 Research on Twitch Platform Channel Recommendation Based on User Usage Mining |
author_sort |
Chih-Yu Chen |
title |
Research on Twitch Platform Channel Recommendation Based on User Usage Mining |
title_short |
Research on Twitch Platform Channel Recommendation Based on User Usage Mining |
title_full |
Research on Twitch Platform Channel Recommendation Based on User Usage Mining |
title_fullStr |
Research on Twitch Platform Channel Recommendation Based on User Usage Mining |
title_full_unstemmed |
Research on Twitch Platform Channel Recommendation Based on User Usage Mining |
title_sort |
research on twitch platform channel recommendation based on user usage mining |
publishDate |
2019 |
url |
http://ndltd.ncl.edu.tw/handle/24a6v4 |
work_keys_str_mv |
AT chihyuchen researchontwitchplatformchannelrecommendationbasedonuserusagemining AT chénchìyǔ researchontwitchplatformchannelrecommendationbasedonuserusagemining AT chihyuchen zhíjīyúguānzhòngxíngwèifēnxīzhītwitchpíngtáipíndàotuījiànxìtǒng AT chénchìyǔ zhíjīyúguānzhòngxíngwèifēnxīzhītwitchpíngtáipíndàotuījiànxìtǒng |
_version_ |
1719257096901361664 |