Probleemi lahendamise ajakulu prognoosimine süvaõppe abil
Failid
Kuupäev
2019
Autorid
Ajakirja pealkiri
Ajakirja ISSN
Köite pealkiri
Kirjastaja
Abstrakt
Probleemi lahendamise ajakulu prognoosimine tarkvaraprojektide korral on suure tähtsusega, kuna selliste projektide planeerimine on raske. Probleemi lahendamisele kuluva aja täpne hindamine on eriti vajalik agiilses tarkvaraarenduses, nt sprindi planeerimises, sest see võimaldab planeerida täpselt. Käesolev magistritöö keskendub antud probleemi kaasaegsetele lähenemisviisidele ja nende efektiivsuse uurimisele. Töös arutletakse selle üle, kuidas on võimalik struktureerida ja rakendada süvaõppe algoritme probleemi lahendamisele kuluva aja prognoosimiseks. Süvaõppe meetodite abil saadud tulemusi võrreldakse teiste kaasaegsete tulemustega. Andmed, mida antud lõputöös kasutatakse, sisaldavad umbes 700 000 ülesannet. Andmed on kollektiivselt kogutud samas valdkonnas varasemalt läbiviidud uurimustest. Kasutades olemasolevaid andmeid, on töös plaanis valideerida olemasolevaid tulemusi ja neid täiendada.
Issue resolution time prediction has a large importance in software projects since planning of these projects are typically hard. Especially in the agile practices, such as sprint planning, to be able to predict correctly how long it would take to resolve an issue, holds the power to plan correctly. This thesis focuses on the state of the art approaches to this problem and study their performances. On top of that we discuss how can one structure and implement a deep learning algorithm to solve issue resolution time prediction problem. Afterwards, we compare and discuss the results of the applied deep learning technique with the current state of the art.The data used for this study contains around 700,000 issues. This data is gathered collectively from the previous studies in this field. By using the already existing data, we plan to validate the existing results and build on top of the current baseline.
Issue resolution time prediction has a large importance in software projects since planning of these projects are typically hard. Especially in the agile practices, such as sprint planning, to be able to predict correctly how long it would take to resolve an issue, holds the power to plan correctly. This thesis focuses on the state of the art approaches to this problem and study their performances. On top of that we discuss how can one structure and implement a deep learning algorithm to solve issue resolution time prediction problem. Afterwards, we compare and discuss the results of the applied deep learning technique with the current state of the art.The data used for this study contains around 700,000 issues. This data is gathered collectively from the previous studies in this field. By using the already existing data, we plan to validate the existing results and build on top of the current baseline.