Method Miningに基づくerror-proneモジュールの予測
Error-prone module prediction using method mining technique

Hideaki UCHIMIYA, Shinpei OGATA, Kenji KAIJIRI
2014 Baiomedikaru, Fajii, Shisutemu Gakkaishi  
Shinshu University , Interdisciplinary Graduate S 酌 00 ' ρブ Seienee (隻 Teehnology . 4ゐ∫かaet ' ln order to predict peJプ -)rmance or ・ reliab 伽 ・ 〔 -ズ0 郷 繊 inelU伽 9 a medica 'i 吻 厂η2磁 0 η ∫ ツ ∫' e 〃1 , prediction 〃let乃 ods using a knowledg ¢ わ ase is often used , and variOUS methods haレe been proρ osed . 〃 b照 レ er, the accura (] y OfpredictiOn depends on the charaCteristics qブ as ア ∫ ' 翩 , SO . finding the best predictor is difificult . 1η ∫ q伽 are engineering domain , for example , in error 一 ρ
more » ... ple , in error 一 ρ rone moduie prediction , this situation ' s the same , SO ' ガ is neceSS α tly tO choose the ( -ρ 伽 al methodfor each target syste 〃1. Zhimin has proposed a selection teehnique of 伽 励 te predictors . ln this paper , 眤 脚 dified Zhit 痂 IS technique in order to P 召 ゆ 厂〃2 〃 lining 〃 lore a ρ pro ρ ri − ate ! y, and } V召 ∫ アro ρ osed the US α ge 〔 -〃 he features qブsystem characteristiCS . Me penyformed three experime 〃 孟 ∫ based on original ( lata and P1-O 層 7SE data, and showed the (コ ffectivene ∬ our proposal . This technique can be aPP -ied・to・variOUS ( -pti 〃 ialpredictor selectionprobiems . KeywordSt metriCS , ぬ ガ α 履 η 勿 9 error ・ prone module , prediction ,ide〃tt : fication , Experimental sq . かare engi 一 η εε厂加 9Hideaki
doi:10.24466/jbfsa.16.2_45 fatcat:asxhprew2bgsne5nz4ktipj4xu