https:\/\/www.youtube.com\/watch?v=YrhMmBeLOJc<\/a><\/p>\n\n\n\nMembres du jury :<\/strong><\/p>\n\n\n\nM. Sylvain LE CORFF<\/strong>, Professeur, Sorbonne Universit\u00e9, FRANCE – Directeur de these
M. Fran\u00e7ois DESBOUVRIES<\/strong>, Professeur, Telecom SudParis, FRANCE – Examinateur
Mme Marie PERROT-DOCKES<\/strong>, Ma\u00eetresse de conf\u00e9rences, Universit\u00e9 de Paris, FRANCE – Examinateur
M. Lionel TRUQUET<\/strong>, Professeur, ENSAI, FRANCE – Rapporteur
M. Joseph RYNKIEWICZ<\/strong>, Ma\u00eetre de conf\u00e9rences, Universit\u00e9 Paris 1, FRANCE – Rapporteur<\/p>\n\n\n\n\u00ab mod\u00e8les de pr\u00e9vision de s\u00e9ries temporelles appliqu\u00e9s \u00e0 de grands ensembles de donn\u00e9es avec inclusion de signaux externes \u00bb<\/h2>\n\n\n\npr\u00e9sent\u00e9 par Monsieur Etienne DAVID<\/h2>\n\n\n\n
R\u00e9sum\u00e9 :<\/strong><\/p>\n\n\n\nLa pr\u00e9vision de s\u00e9ries temporelles est un probl\u00e8me math\u00e9matique r\u00e9pandu dans de nombreux secteurs, devenant un v\u00e9ritable d\u00e9fi pour les m\u00e9thodes existantes de la litt\u00e9rature lorsque de grands ensembles de donn\u00e9es rassemblant des milliers de s\u00e9ries temporelles et des signaux externes sont consid\u00e9r\u00e9s. Une illustration concr\u00e8te de ce probl\u00e8me peut \u00eatre trouv\u00e9e dans l\u2019industrie de la mode o\u00f9 ses acteurs tentent d\u2019anticiper l\u2019\u00e9volution de milliers de v\u00eatements pour cr\u00e9er leurs collections, analysant les comportements des influenceurs pour proposer la mode de demain. En utilisant cette application comme fil conducteur, nous pr\u00e9sentons trois contributions explorant diff\u00e9rentes r\u00e9ponses concernant le probl\u00e8me de pr\u00e9vision de s\u00e9ries temporelles o\u00f9 de grands ensembles de donn\u00e9es et des signaux externes sont consid\u00e9r\u00e9s. Une premi\u00e8re r\u00e9ponse est propos\u00e9e avec l’introduction d’un nouveau mod\u00e8le hybride et la publication d’un large ensemble de donn\u00e9es rassemblant 10000 s\u00e9ries temporelles de mode et des signaux externes d’influenceurs. Une seconde approche est ensuite \u00e9tudi\u00e9e avec un travail th\u00e9orique sur les mod\u00e8les de Markov cach\u00e9s \u00e0 signaux externes. Enfin, une derni\u00e8re r\u00e9ponse est propos\u00e9e avec l’introduction d’une nouvelle m\u00e9thode m\u00e9langeant le fonctionnement interne des mod\u00e8les de Markov cach\u00e9s avec des r\u00e9seaux de neurones. Les r\u00e9sultats pr\u00e9sent\u00e9s dans ces trois contributions ont mis en \u00e9vidence plusieurs \u00e9l\u00e9ments de r\u00e9ponse. Premi\u00e8rement, les r\u00e9seaux de neurones sont d\u00e9cisifs pour traiter de grands ensembles de donn\u00e9es et sont particuli\u00e8rement bien con\u00e7us pour exploiter des signaux externes. Deuxi\u00e8mement, les mod\u00e8les de Markov cach\u00e9s avec signaux externes sont \u00e9galement des m\u00e9thodes efficaces, capables de capturer des d\u00e9pendances complexes entre des s\u00e9ries temporelles et leurs signaux externes. Cependant, ils ne parviennent pas \u00e0 g\u00e9rer de grands ensembles de donn\u00e9es car un mod\u00e8le doit \u00eatre entra\u00een\u00e9 pour chaque nouvelle s\u00e9rie temporelle. Enfin, inspir\u00e9s par les r\u00e9sultats frappants des mod\u00e8les de Markov cach\u00e9s avec des signaux externes, nous montrons que l’introduction de processus cach\u00e9s dans des mod\u00e8les bas\u00e9s sur des r\u00e9seaux neuronaux peut les aider \u00e0 explorer plus profond\u00e9ment les grands ensembles de donn\u00e9es, \u00e0 mod\u00e9liser une plus grande vari\u00e9t\u00e9 de comportements et \u00e0 exploiter plus finement les signaux externes.
Abstract :<\/strong><\/p>\n\n\n\nTime series forecasting is a widespread mathematical problem in numerous sectors becoming a real challenge for existing methods of the literature where large datasets gathering thousands of time series and external signals are considered. A concrete illustration of this issue can be find in the fashion industry where its actors try to anticipate the evolution of thousands of garments to create their collections, analysing influencers and early adopters behaviours to propose the fashion of tomorrow. Using this application as a common thread, we present three contributions exploring different answers regarding the time series forecasting problem where large datasets and external signals are considered. A first answer is proposed with the introduction of a new hybrid model and the publication of a large dataset gathering 10000 fashion time series and influencers external signals. A second approach is then studied with theoretical work done on hidden Markov models with external signals. Finally, a last answer is proposed with the introduction of a new method mixing the inner workings of hidden Markov model and neural networks. Results presented in this three contribution highlighted several elements of answer. Firstly, neural networks are decisive to deal with large datasets and they are particularly well designed to leverage external signals. Secondly, hidden Markov models with external signals are also strong methods that can capture complex dependencies between time series and their external signals. However, they fail at handling large datasets as a model has to be trained for each new time series. Finally, inspired by the striking results of hidden Markov models with external signals, we reveal that introducing hidden processes in neural-network-based models can help them explore large datasets more deeply, model a richer variety of behaviour and leverage more finely external signals.<\/p>\n","protected":false},"excerpt":{"rendered":"
L’Ecole doctorale : Math\u00e9matiques Hadamard et le Laboratoire de recherche SAMOVAR – Services r\u00e9partis, Architectures, Mod\u00e9lisation, Validation, Administration des R\u00e9seaux pr\u00e9sentent l\u2019AVIS DE SOUTENANCE de Monsieur Etienne DAVID Autoris\u00e9 \u00e0 pr\u00e9senter ses travaux en vue de l\u2019obtention du Doctorat de l’Institut Polytechnique de Paris, pr\u00e9par\u00e9 \u00e0 T\u00e9l\u00e9com SudParis en : \u00ab mod\u00e8les de pr\u00e9vision de […]<\/p>\n","protected":false},"author":4,"featured_media":0,"comment_status":"closed","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"ocean_post_layout":"","ocean_both_sidebars_style":"","ocean_both_sidebars_content_width":0,"ocean_both_sidebars_sidebars_width":0,"ocean_sidebar":"","ocean_second_sidebar":"","ocean_disable_margins":"enable","ocean_add_body_class":"","ocean_shortcode_before_top_bar":"","ocean_shortcode_after_top_bar":"","ocean_shortcode_before_header":"","ocean_shortcode_after_header":"","ocean_has_shortcode":"","ocean_shortcode_after_title":"","ocean_shortcode_before_footer_widgets":"","ocean_shortcode_after_footer_widgets":"","ocean_shortcode_before_footer_bottom":"","ocean_shortcode_after_footer_bottom":"","ocean_display_top_bar":"default","ocean_display_header":"default","ocean_header_style":"","ocean_center_header_left_menu":"","ocean_custom_header_template":"","ocean_custom_logo":0,"ocean_custom_retina_logo":0,"ocean_custom_logo_max_width":0,"ocean_custom_logo_tablet_max_width":0,"ocean_custom_logo_mobile_max_width":0,"ocean_custom_logo_max_height":0,"ocean_custom_logo_tablet_max_height":0,"ocean_custom_logo_mobile_max_height":0,"ocean_header_custom_menu":"","ocean_menu_typo_font_family":"","ocean_menu_typo_font_subset":"","ocean_menu_typo_font_size":0,"ocean_menu_typo_font_size_tablet":0,"ocean_menu_typo_font_size_mobile":0,"ocean_menu_typo_font_size_unit":"px","ocean_menu_typo_font_weight":"","ocean_menu_typo_font_weight_tablet":"","ocean_menu_typo_font_weight_mobile":"","ocean_menu_typo_transform":"","ocean_menu_typo_transform_tablet":"","ocean_menu_typo_transform_mobile":"","ocean_menu_typo_line_height":0,"ocean_menu_typo_line_height_tablet":0,"ocean_menu_typo_line_height_mobile":0,"ocean_menu_typo_line_height_unit":"","ocean_menu_typo_spacing":0,"ocean_menu_typo_spacing_tablet":0,"ocean_menu_typo_spacing_mobile":0,"ocean_menu_typo_spacing_unit":"","ocean_menu_link_color":"","ocean_menu_link_color_hover":"","ocean_menu_link_color_active":"","ocean_menu_link_background":"","ocean_menu_link_hover_background":"","ocean_menu_link_active_background":"","ocean_menu_social_links_bg":"","ocean_menu_social_hover_links_bg":"","ocean_menu_social_links_color":"","ocean_menu_social_hover_links_color":"","ocean_disable_title":"default","ocean_disable_heading":"default","ocean_post_title":"","ocean_post_subheading":"","ocean_post_title_style":"","ocean_post_title_background_color":"","ocean_post_title_background":0,"ocean_post_title_bg_image_position":"","ocean_post_title_bg_image_attachment":"","ocean_post_title_bg_image_repeat":"","ocean_post_title_bg_image_size":"","ocean_post_title_height":0,"ocean_post_title_bg_overlay":0.5,"ocean_post_title_bg_overlay_color":"","ocean_disable_breadcrumbs":"default","ocean_breadcrumbs_color":"","ocean_breadcrumbs_separator_color":"","ocean_breadcrumbs_links_color":"","ocean_breadcrumbs_links_hover_color":"","ocean_display_footer_widgets":"default","ocean_display_footer_bottom":"default","ocean_custom_footer_template":"","ocean_post_oembed":"","ocean_post_self_hosted_media":"","ocean_post_video_embed":"","ocean_link_format":"","ocean_link_format_target":"self","ocean_quote_format":"","ocean_quote_format_link":"post","ocean_gallery_link_images":"on","ocean_gallery_id":[],"footnotes":""},"categories":[286,615],"tags":[],"class_list":["post-6458","post","type-post","status-publish","format-standard","hentry","category-fractualites-ennews-fr","category-seminaire-sop","entry"],"_links":{"self":[{"href":"https:\/\/samovar.telecom-sudparis.eu\/index.php\/wp-json\/wp\/v2\/posts\/6458","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/samovar.telecom-sudparis.eu\/index.php\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/samovar.telecom-sudparis.eu\/index.php\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/samovar.telecom-sudparis.eu\/index.php\/wp-json\/wp\/v2\/users\/4"}],"replies":[{"embeddable":true,"href":"https:\/\/samovar.telecom-sudparis.eu\/index.php\/wp-json\/wp\/v2\/comments?post=6458"}],"version-history":[{"count":1,"href":"https:\/\/samovar.telecom-sudparis.eu\/index.php\/wp-json\/wp\/v2\/posts\/6458\/revisions"}],"predecessor-version":[{"id":6459,"href":"https:\/\/samovar.telecom-sudparis.eu\/index.php\/wp-json\/wp\/v2\/posts\/6458\/revisions\/6459"}],"wp:attachment":[{"href":"https:\/\/samovar.telecom-sudparis.eu\/index.php\/wp-json\/wp\/v2\/media?parent=6458"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/samovar.telecom-sudparis.eu\/index.php\/wp-json\/wp\/v2\/categories?post=6458"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/samovar.telecom-sudparis.eu\/index.php\/wp-json\/wp\/v2\/tags?post=6458"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}