講演名 2018-03-08
Banner Click Through Rate Classification Using Deep Neural Convolutional Network
Nicolas Michel(東大), Hayato Sakata(SMN), Keita Kurita(東大), Toshihiko Yamasaki(東大),
PDFダウンロードページ PDFダウンロードページへ
抄録(和) In banner advertising, Click Through Rate (CTR) is one of the most important indicators to evaluate oneadvertisement?s quality. Advertisers create massive number of banner candidates in empirical ways, then proceed to actual testsby delivering advertisement to measure each banner?s effectiveness. This process is expensive and therefore our CTR predictionhelps reducing online advertising costs. In this work, we propose a method to classify effective and ineffective advertisingbanners based on image processing using state-of-the-art CNN. We first focus only on images then conduct experiments includingmetadata (product, advertiser, etc) to increase the CTR prediction accuracy and demonstrate which metadata is the mostinfluential. Subsequently, each approach is compared to human performance. In the second part of our work, we detect whichparts of the image contribute predominantly to increase the CTR by generating heat maps for each classes. This work leads to adeeper understanding of a banner advertising success and helps making decisions on how to improve it.
抄録(英) In banner advertising, Click Through Rate (CTR) is one of the most important indicators to evaluate oneadvertisement?s quality. Advertisers create massive number of banner candidates in empirical ways, then proceed to actual testsby delivering advertisement to measure each banner?s effectiveness. This process is expensive and therefore our CTR predictionhelps reducing online advertising costs. In this work, we propose a method to classify effective and ineffective advertisingbanners based on image processing using state-of-the-art CNN. We first focus only on images then conduct experiments includingmetadata (product, advertiser, etc) to increase the CTR prediction accuracy and demonstrate which metadata is the mostinfluential. Subsequently, each approach is compared to human performance. In the second part of our work, we detect whichparts of the image contribute predominantly to increase the CTR by generating heat maps for each classes. This work leads to adeeper understanding of a banner advertising success and helps making decisions on how to improve it.
キーワード(和) Convolutional Neural Network / Click Through Rate / Deep Learning / Banner Advertising / Deep Learning
キーワード(英) Convolutional Neural Network / Click Through Rate / Deep Learning / Banner Advertising / Deep Learning
資料番号 IMQ2017-43,IE2017-135,MVE2017-85
発行日 2018-03-01 (IMQ, IE, MVE)

研究会情報
研究会 CQ / MVE / IE / IMQ
開催期間 2018/3/8(から2日開催)
開催地(和) 沖縄産業支援センター
開催地(英) Okinawa Industry Support Center
テーマ(和) 五感メディア,食メディア,マルチメディア, メディアエクスペリエンス,映像符号化, イメージメディアの品質,ネットワークの品質および信頼性,一般(食メディア(CEA)研究会,魅力工学(AC)研究会協賛)
テーマ(英) Five Senses Media, Cooking and Eating Activities Media, Multimedia, Media Experience, Video Encoding, Image Media Quality, Network Quality and Reliability, etc. (Co-sponsor: Technical Committee on Multimedia on Cooking and Eating Activities (CEA))
委員長氏名(和) 林 孝典(広島工大) / 亀田 能成(筑波大) / 浜本 隆之(東京理科大) / 杉山 賢二(成蹊大)
委員長氏名(英) Takanori Hayashi(Hiroshima Inst. of Tech.) / Yoshinari Kameda(Univ. of Tsukuba) / Takayuki Hamamoto(Tokyo Univ. of Science) / Kenji Sugiyama(Seikei Univ.)
副委員長氏名(和) 下西 英之(NEC) / 岡本 淳(NTT) / 間瀬 健二(名大) / 児玉 和也(NII) / 木全 英明(NTT) / 中口 俊哉(千葉大) / 前田 充(キヤノン)
副委員長氏名(英) Hideyuki Shimonishi(NEC) / Jun Okamoto(NTT) / Kenji Mase(Nagoya Univ.) / Kazuya Kodama(NII) / Hideaki Kimata(NTT) / Toshiya Nakaguchi(Chiba Univ.) / Mitsuru Maeda(Canon)
幹事氏名(和) 池上 大介(NTT) / 久保 亮吾(慶大) / 飯山 将晃(京大) / 木村 篤信(NTT) / 内山 英昭(九大) / 高橋 桂太(名大) / 河村 圭(KDDI総合研究所) / 工藤 博章(名大) / 齊藤 新一郎(ソニー)
幹事氏名(英) Daisuke Ikegami(NTT) / Ryogo Kubo(Keio Univ.) / Masaaki Iiyama(Kyoto Univ.) / Atsunobu Kimura(NTT) / Hideaki Uchiyama(Kyushu Univ.) / Keita Takahashi(Nagoya Univ.) / Kei Kawamura(KDDI Research) / Hiroaki Kudo(Nagoya Univ.) / Shinichiro Saito(Sony)
幹事補佐氏名(和) 大田 健紘(日本工大) / 福元 徳広(KDDI総合研究所) / 山本 嶺(電通大) / 平山 高嗣(名大) / 青木 良輔(NTT) / 松尾 康孝(NHK) / 早瀬 和也(NTT) / 土田 勝(NTT) / 大橋 剛介(静岡大)
幹事補佐氏名(英) Kenko Ota(Nippon Inst. of Tech.) / Norihiro Fukumoto(KDDI Research Inc.) / Ryo Yamamoto(UEC) / Takatsugu Hirayama(Nagoya Univ.) / Ryosuke Aoki(NTT) / Yasutaka Matsuo(NHK) / Kazuya Hayase(NTT) / Masaru Tsuchida(NTT) / Gosuke Ohashi(Shizuoka Univ.)

講演論文情報詳細
申込み研究会 Technical Committee on Communication Quality / Technical Committee on Media Experience and Virtual Environment / Technical Committee on Image Engineering / Technical Committee on Image Media Quality
本文の言語 ENG
タイトル(和)
サブタイトル(和)
タイトル(英) Banner Click Through Rate Classification Using Deep Neural Convolutional Network
サブタイトル(和)
キーワード(1)(和/英) Convolutional Neural Network / Convolutional Neural Network
キーワード(2)(和/英) Click Through Rate / Click Through Rate
キーワード(3)(和/英) Deep Learning / Deep Learning
キーワード(4)(和/英) Banner Advertising / Banner Advertising
キーワード(5)(和/英) Deep Learning / Deep Learning
第 1 著者 氏名(和/英) Nicolas Michel / Nicolas Michel
第 1 著者 所属(和/英) University Of Tokyo(略称:東大)
University Of Tokyo(略称:Univ. Of Tokyo)
第 2 著者 氏名(和/英) Hayato Sakata / Hayato Sakata
第 2 著者 所属(和/英) So-net Media Networks(略称:SMN)
So-net Media Networks(略称:SMN)
第 3 著者 氏名(和/英) Keita Kurita / Keita Kurita
第 3 著者 所属(和/英) University Of Tokyo(略称:東大)
University Of Tokyo(略称:Univ. Of Tokyo)
第 4 著者 氏名(和/英) Toshihiko Yamasaki / Toshihiko Yamasaki
第 4 著者 所属(和/英) University Of Tokyo(略称:東大)
University Of Tokyo(略称:Univ. Of Tokyo)
発表年月日 2018-03-08
資料番号 IMQ2017-43,IE2017-135,MVE2017-85
巻番号(vol) vol.117
号番号(no) IMQ-483,IE-484,MVE-485
ページ範囲 pp.101-106(IMQ), pp.101-106(IE), pp.101-106(MVE),
ページ数 6
発行日 2018-03-01 (IMQ, IE, MVE)