{"created":"2025-01-19T01:10:20.470242+00:00","updated":"2025-01-19T15:04:07.062020+00:00","metadata":{"_oai":{"id":"oai:ipsj.ixsq.nii.ac.jp:00208999","sets":["581:10433:10434"]},"path":["10434"],"owner":"44499","recid":"208999","title":["深層学習による時間減衰を考慮したインフィード広告のCTR予測"],"pubdate":{"attribute_name":"公開日","attribute_value":"2021-01-15"},"_buckets":{"deposit":"0e2fd33f-efd5-4bb2-8990-b10bca0ee65e"},"_deposit":{"id":"208999","pid":{"type":"depid","value":"208999","revision_id":0},"owners":[44499],"status":"published","created_by":44499},"item_title":"深層学習による時間減衰を考慮したインフィード広告のCTR予測","author_link":["525778","525779","525781","525777","525780","525782"],"item_titles":{"attribute_name":"タイトル","attribute_value_mlt":[{"subitem_title":"深層学習による時間減衰を考慮したインフィード広告のCTR予測"},{"subitem_title":"Deep Time-decaying CTR Prediction of In-feed Advertising","subitem_title_language":"en"}]},"item_keyword":{"attribute_name":"キーワード","attribute_value_mlt":[{"subitem_subject":"[特集：5G時代の社会を創るモバイル・高度交通システム] 深層学習，マルチモーダル学習，インターネット広告，クリック率予測","subitem_subject_scheme":"Other"}]},"item_type_id":"2","publish_date":"2021-01-15","item_2_text_3":{"attribute_name":"著者所属","attribute_value_mlt":[{"subitem_text_value":"株式会社NTTドコモ／大阪大学"},{"subitem_text_value":"株式会社NTTドコモ"},{"subitem_text_value":"大阪大学"}]},"item_2_text_4":{"attribute_name":"著者所属(英)","attribute_value_mlt":[{"subitem_text_value":"NTT DOCOMO, INC. / Osaka University","subitem_text_language":"en"},{"subitem_text_value":"NTT DOCOMO, INC.","subitem_text_language":"en"},{"subitem_text_value":"Osaka University","subitem_text_language":"en"}]},"item_language":{"attribute_name":"言語","attribute_value_mlt":[{"subitem_language":"jpn"}]},"publish_status":"0","weko_shared_id":-1,"item_file_price":{"attribute_name":"Billing file","attribute_type":"file","attribute_value_mlt":[{"url":{"url":"https://ipsj.ixsq.nii.ac.jp/record/208999/files/IPSJ-JNL6201036.pdf","label":"IPSJ-JNL6201036.pdf"},"date":[{"dateType":"Available","dateValue":"2023-01-15"}],"format":"application/pdf","billing":["billing_file"],"filename":"IPSJ-JNL6201036.pdf","filesize":[{"value":"936.8 kB"}],"mimetype":"application/pdf","priceinfo":[{"tax":["include_tax"],"price":"660","billingrole":"5"},{"tax":["include_tax"],"price":"330","billingrole":"6"},{"tax":["include_tax"],"price":"0","billingrole":"8"},{"tax":["include_tax"],"price":"0","billingrole":"44"}],"accessrole":"open_date","version_id":"ddc8ae19-6793-4776-8533-ad010095a6d4","displaytype":"detail","licensetype":"license_note","license_note":"Copyright (c) 2021 by the Information Processing Society of Japan"}]},"item_2_creator_5":{"attribute_name":"著者名","attribute_type":"creator","attribute_value_mlt":[{"creatorNames":[{"creatorName":"出水, 宰"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"深澤, 佑介"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"森田, 浩"}],"nameIdentifiers":[{}]}]},"item_2_creator_6":{"attribute_name":"著者名(英)","attribute_type":"creator","attribute_value_mlt":[{"creatorNames":[{"creatorName":"Tsukasa, Demizu","creatorNameLang":"en"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"Yusuke, Fukazawa","creatorNameLang":"en"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"Hiroshi, Morita","creatorNameLang":"en"}],"nameIdentifiers":[{}]}]},"item_2_source_id_9":{"attribute_name":"書誌レコードID","attribute_value_mlt":[{"subitem_source_identifier":"AN00116647","subitem_source_identifier_type":"NCID"}]},"item_resource_type":{"attribute_name":"資源タイプ","attribute_value_mlt":[{"resourceuri":"http://purl.org/coar/resource_type/c_6501","resourcetype":"journal article"}]},"item_2_source_id_11":{"attribute_name":"ISSN","attribute_value_mlt":[{"subitem_source_identifier":"1882-7764","subitem_source_identifier_type":"ISSN"}]},"item_2_description_7":{"attribute_name":"論文抄録","attribute_value_mlt":[{"subitem_description":"インフィード広告は，ユーザへの視認性が高いため，クリック率（CTR）がより高い広告クリエイティブを表示することが重要になる．しかし，インフィード広告はその高頻度な表示のために，配信以後のCTRの時間的な減衰が速いという特徴がある．そこで本研究では，この時間減衰を考慮したうえでのCTR予測手法を提案する．まず，広告クリエイティブの画像情報やテキスト情報，配信設定情報といったマルチモーダルな特徴量からCTRをロバストに予測するモデルを構築する．次に，CTRの時系列変化を抽象的に表現可能なRNNモデルを構築する．アドネットワーク上の配信データを用いたオフラインでの多期間のCTR予測検証を行い，提案手法の有効性を示す．","subitem_description_type":"Other"}]},"item_2_description_8":{"attribute_name":"論文抄録(英)","attribute_value_mlt":[{"subitem_description":"In-feed advertisement are characterized by high visibility to users, so it is important to display creatives with higher click-through rate (CTR). However, since in-feed advertisements are displayed to users frequently, the CTR has a rapid time-decay after the advertisement is delivered. In this study, we propose a CTR prediction method considering its time-decay for in-feed advertisements. First, we introduce a model can robustly predict CTR from multi-modal features such as image information, text information and audience setting information of advertising creatives. Next, we construct an RNN model considering the time series change of CTR. The effectiveness of our proposed model is demonstrated by offline verification of multi-period CTR prediction task using historical and contextual data on an ad network in Japan.","subitem_description_type":"Other"}]},"item_2_biblio_info_10":{"attribute_name":"書誌情報","attribute_value_mlt":[{"bibliographicPageEnd":"301","bibliographic_titles":[{"bibliographic_title":"情報処理学会論文誌"}],"bibliographicPageStart":"292","bibliographicIssueDates":{"bibliographicIssueDate":"2021-01-15","bibliographicIssueDateType":"Issued"},"bibliographicIssueNumber":"1","bibliographicVolumeNumber":"62"}]},"relation_version_is_last":true,"item_2_identifier_registration":{"attribute_name":"ID登録","attribute_value_mlt":[{"subitem_identifier_reg_text":"10.20729/00208897","subitem_identifier_reg_type":"JaLC"}]},"weko_creator_id":"44499"},"id":208999,"links":{}}