izpis_h1_title_alt

Napovedovanje časovnih vrst z nevronskimi mrežami z dolgim kratkoročnim spominom
Roštan, Teja (Author), Guid, Matej (Mentor) More about this mentor... This link opens in a new window, Vodopivec, Tom (Co-mentor)

.pdfPDF - Presentation file, Download (7,13 MB)

Abstract
Za napovedovanje časovnih vrst je dolgo veljalo načelo, da enostavne metode v napovednih točnostih presegajo metode strojnega učenja. Vendar pa enostavni modeli ne znajo izrabljati raznovrstnih medsebojnih odvisnosti in informacij, ki jih ponujajo časovne vrste, vsebinsko sorodne ali podobne tistim, ki so predmet napovedi. Pojav masovnih podatkov je povezan tudi z zbiranjem ogromnega števila časovnih vrst, vendar pa ob uporabi enostavnih, klasičnih metod, njihov visok potencial za izboljšanje natančnosti napovedi ostaja neizkoriščen. Nevronske mreže so dobile priložnost, da zapolnijo omenjeno vrzel. Za delo z zaporednimi podatki so primerne povratne nevronske mreže, ki pri napovedih znajo izkoriščati medsebojne odvisnosti časovnih točk. Med njimi veljajo za še zlasti uspešne pri napovedovanju časovnih vrst tako imenovane nevronske mreže z dolgim kratkoročnim spominom. V delu smo se osredotočili na izgradnjo in optimizacijo tega tipa nevronskih mrež. Naš namen je bil izboljšati napovedno točnost pri napovedovanju časovnih vrst ter hkrati razumeti, zakaj in koliko k temu izboljšanju prispevajo posamezni dejavniki. Napovedovali smo število klikov na oglase na družabnem omrežju Facebook. Najprej smo analizirali različne kombinacije obdelav časovnih vrst, za katere se je izkazalo, da bi lahko vplivale na izboljšanje natančnosti napovedi. Nevronske mreže smo učili na skupini sorodnih časovnih vrst in napovedi primerjali s klasičnimi pristopi napovedovanja časovnih vrst ARIMA, ARIMAX in VAR. Raziskali smo tudi več možnosti za izboljšanje uspešnosti napovedovanja z nevronskimi mrežami s pomočjo uporabe podobnih časovnih vrst. Po pričakovanjih se je izkazalo, da nevronske mreže z dolgim kratkoročnim spominom ob ustrezni obdelavi podatkov dosegajo višjo napovedno točnost kot klasični modeli. Pokazali smo, da je z uporabo podobnih časovnih vrst napovedi možno še dodatno izboljšati, vendar pa ta pristop vselej ne pomaga.

Language:Slovenian
Keywords:analiza časovnih vrst, obdelava časovnih vrst, napovedovanje, nevronske mreže z dolgim kratkoročnim spominom, spletno oglaševanje, gručenje
Work type:Master's thesis/paper (mb22)
Organization:FRI - Faculty of computer and information science
Year:2018
Views:292
Downloads:342
Metadata:XML RDF-CHPDL DC-XML DC-RDF
 
Average score:(0 votes)
Your score:Voting is allowed only to logged in users.
:
Share: Bookmark and Share

Secondary language

Language:English
Title:Time series forecasting with long short-term memory neural networks
Abstract:
Time series forecasting was for a long time based on the principle that simple methods in forecast accuracy exceed machine learning methods. However, simple models cannot use the various mutual dependencies and information offered by time series, content-related or similar to those that are subject to prediction. The occurrence of massive data is also connected with the collecting of enormous amounts of time series, but in using simple, traditional methods, their high potential for enhancing forecast accuracy remains untapped. The opportunity to bridge this gap comes with neural networks. To process sequential data, recurrent neural networks are used in forecasting that can use mutual dependencies of points in time. Among recurrent neural networks, long short-term memory neural networks are considered as especially successful in time series forecasting. The paper focuses on the building and optimization of this neural network type. Our purpose was to improve forecast accuracy in time series forecasting, understanding at the same time why and to what degree individual factors contribute to this improvement. The forecasting was applied to the number of clicks on Facebook ads. First, we analysed various combinations of time series processing, discovering that they might influence forecast accuracy improvements. Neural networks learned using a group of related time series and the forecasts were compared to the traditional time series forecasting approaches ARIMA, ARIMAX and VAR. We also researched a number of options to improve forecast effectiveness with neural networks using similar time series. As expected, we found that with adequate data processing, long short-term memory neural networks achieve greater forecast accuracy compared to the traditional models. We demonstrated that forecasting can be further improved using similar time series, however, this approach is not always helpful.

Keywords:time series analysis, time series processing, forecasting, long short-term memory neural networks, digital advertising, clustering

Similar documents

Similar works from RUL:
Similar works from other Slovenian collections:

Comments

Leave comment

You have to log in to leave a comment.

Comments (0)
0 - 0 / 0
 
There are no comments!

Back