doktorska disertacija
Danijel Radjenović (Author), Aleš Živkovič (Mentor), Richard Torkar (Co-mentor)

Abstract

Metrike programske opreme so uporabljene v modelih za napovedovanje napak programske opreme. Model določi dele izvorne kode, ki vsebujejo napake in jih je potrebno pregledati. S tem usmerja proces zagotavljanja kakovosti programske opreme. Z identifikacijo napak se izboljša kakovost programske opreme, z določanjem njihove lokacije pa se znižajo stroški testiranja. Metoderazvoja programske opreme se počasi oddaljujejo od tradicionalnega modela slapu (Waterfall) k bolj agilnim pristopom (Scrum, XP), s tem pa se spreminja tudi način testiranja programske opreme. V agilnih okoljih si ne moremo privoščiti, da bi izvedli testiranje celotnega sistema za vsako različico programske opreme, zato testiramo le dele, ki jih je napovedovalni model določil za nepravilne. Glavni namen je bil razviti učinkovit model za napovedovanje napak programske opreme v agilnih okoljih. Da smo model razvili,smo opravili sistematični pregled literature, analizirali produktne inprocesne metrike ter ocenili klasifikacijske tehnike. Dobljeni model smo ovrednotili v industrijskem okolju. Model, ki vsebuje zgolj štiri procesne metrike, se je izkazal za uspešnega v industrijskem okolju, saj je odkril več kot polovico vseh napak (popolnost > 0,5) in ni vseboval enkrat več lažno pozitivnih napovedi (pravilnost > 0,33). Zaradi visoke multikolinearnosti metrik je bil uspešnejši od polnega modela, ki vsebuje 80 metrik. Procesne metrike so bile uspešne pri napovedovanju napak, medtem ko produktne metrike niso bile. Napake v agilnih okoljih so najbolj pogojene s starostjo in velikostjo sprememb programske opreme, pri čemer imajo nedavne in velike spremembe večjo verjetnost, da vsebujejo napake.

Keywords

metrike programske opreme;napovedovalni modeli;zagotavljanje kakovosti;testiranje;kakovost programske opreme;

Data

Language: Slovenian
Year of publishing:
Typology: 2.08 - Doctoral Dissertation
Organization: UM FERI - Faculty of Electrical Engineering and Computer Science
Publisher: [D. Radjenović]
UDC: 004.416.2:004.89(043.3)
COBISS: 270541568 Link will open in a new window
Views: 1737
Downloads: 220
Average score: 0 (0 votes)
Metadata: JSON JSON-RDF JSON-LD TURTLE N-TRIPLES XML RDFA MICRODATA DC-XML DC-RDF RDF

Other data

Secondary language: English
Secondary title: Software fault prediction framework for agile environments
Secondary abstract: Software metrics are used in models for software fault prediction. The model determines the parts of the source code which contain faults and should be reviewed. It directs the software quality assurance process. By identifying faults, software quality is improved; whereas, by determining their location, costs of testing are reduced. Software development methods are slowly moving away from the traditional Waterfall model to more agile approaches (Scrum, XP), thereby changing the way software testing is performed. In agile environments, we cannot afford to test the entire system for each version of the software; therefore, only the parts determined faulty by the predictive model are tested. The main objective was to develop an effective model for predicting faults in agile environments. In order to develop a model, we have conducted a systematic literature review, analyzed product and process metricsand evaluated classification techniques. The resulting model was evaluated in an industrial environment. The model containing only four processmetrics has proven to be successful in an industrial environment, sinceit discovered more than half of the faults (completeness > 0.5) and contained less than twice as many false positive predictions (correctness > 0,33). Due to the high degree of multicollinearity between metrics, it was more successful than the full model containing 80 metrics. Process metrics were successful in predicting faults, while product metrics were not. Faults in agile environment are most influenced by age and size of software changes, where the recent and significant changes are more likely to contain faults.
Secondary keywords: software metrics;software fault prediction;prediction model;software quality assurance;testing;Računalniška programska oprema;Disertacije;Napake;Napovedovanje;
URN: URN:SI:UM:
Type (COBISS): Doctoral dissertation
Thesis comment: Univ. v Mariboru, Fak. za elektrotehniko, računalništvo in informatiko
Pages: 135 str.
ID: 8727238