Citation Context

...lass cohesion measures under considerationsClass CohesionsMeasuresDefinition/FormulasThe lack of Cohesion insMethods (LCOM1) [36]sLCOM1= Number of pairs of methods that do not share attributes.sLCOM2 =-=[37]-=- P= Number of pairs of methods that do not share attributes.sQ= Number of pairs of methods that share attributes.s otherwise QPifQP LCOM 0 0 2sLCOM3 [38] LCOM3= Number of connected componen...

Citation Context

...ution to the prediction model for the external quality attribute undersconsideration.sClass cohesion measurement aims at quantifying the degree of relatedness of the methodssand attributes in a class =-=[3]-=-. A class can contain methods with different characteristics.sFor example, on one extreme, a constructor typically initializes most or all attributes insthe class, whereas a destructor deinitializes m...

Citation Context

...imilar empirical studies, which allows us to compare our results with thosespreviously obtained. The selected cohesion measures have been well studied from bothstheoretical and empirical perspectives =-=[10, 11, 12, 13, 14, 15, 16]-=-.s2.2. Relations Between Cohesion and SPssThe original definitions of a few of the measures investigated here consider whether tosconsider or ignore SPs in cohesion measurement. Bieman and Kang [4] su...

Citation Context

... number of pairssof methods. The similarity between methods i and j is defined assl II II II jiSimilarity ji ji ji ),min( ),( , where l is the number of attributes.sTight Class Cohesions(TCC) =-=[4]-=-sTCC= The relative number of directly connected pairs of methods, where two methods are directlysconnected if they are directly connected to an attribute. A method m is directly connected to an attrib...

Citation Context

... obtained using each of the prediction models built for eachsconsidered measures in each scenario. None of them found to follow the normalsdistribution. Therefore, we applied the Wilcoxon paired test =-=[34]-=-, which is a standardsnon-parametric statistical technique that is applied to compare two independent sets ofssamples that are not necessarily normal. Based on Wilcoxon test and according to thestypic...

Citation Context

...llection ProcesssWe applied logistic regression (LR) [25], a statistical technique that is widely used tosbuild prediction models for external quality attributes, such as class fault-proneness (e.g.,s=-=[3, 16, 26, 27]-=-), class reuse-proneness (e.g., [22]), and class maintainability (e.g., [28]).sApplying other statistical techniques to construct prediction models, such as those usedsby Briand and Wuest [29], Subram...

Citation Context

...ass maintainability (e.g., [28]).sApplying other statistical techniques to construct prediction models, such as those usedsby Briand and Wuest [29], Subramanyam and Krishnan [30], and Arisholm et al. =-=[31]-=-,sand comparing the results with the LR results are outside the scope of this paper.sIn LR analysis, the variables are classified as dependent or independent. The independentsvariables (i.e., the cons...

Citation Context

...imilar empirical studies, which allows us to compare our results with thosespreviously obtained. The selected cohesion measures have been well studied from bothstheoretical and empirical perspectives =-=[10, 11, 12, 13, 14, 15, 16]-=-.s2.2. Relations Between Cohesion and SPssThe original definitions of a few of the measures investigated here consider whether tosconsider or ignore SPs in cohesion measurement. Bieman and Kang [4] su...

Citation Context

... 16, 26, 27]), class reuse-proneness (e.g., [22]), and class maintainability (e.g., [28]).sApplying other statistical techniques to construct prediction models, such as those usedsby Briand and Wuest =-=[29]-=-, Subramanyam and Krishnan [30], and Arisholm et al. [31],sand comparing the results with the LR results are outside the scope of this paper.sIn LR analysis, the variables are classified as dependent ...

Citation Context

...strated by examples that the LCOM2 (LCOM3) value either decreases (increases)sor remains the same. Without justifying or studying the impact of their choice on the DCDsand DCI values, Badri and Badri =-=[17]-=- suggested accounting for all public methods, whichstypically include all SPs. Chae et al. [5] suggested ignoring SPs when applying CBMCsbecause SPs do not contribute to class cohesiveness but potenti...

Citation Context

...ribute using asprediction model that incorporates the same cohesion measure.sSeveral researchers have qualitatively addressed the problem of considering or ignoringsSPs in cohesion measurement (e.g., =-=[3, 4, 5, 6, 7]-=-). To our knowledge, only Al Dallal [8]shas empirically studied the effect of considering or ignoring SPs. In that study, a group ofsexisting measures was selected and SPs were empirically investigate...

Citation Context

... otherwise QPifQP LCOM 0 0 2sLCOM3 [38] LCOM3= Number of connected components in the graph that represents each method as a node and thessharing of at least one attribute as an edge.sLCOM4 =-=[39]-=- Similar to LCOM3, where additional edges are used to represent method invocations.sLCOM5 [40] LCOM5= (kl-a)/(kl-l), where l is the number of attributes, k is the number of methods, and a is the sumso...

Citation Context

...imilar empirical studies, which allows us to compare our results with thosespreviously obtained. The selected cohesion measures have been well studied from bothstheoretical and empirical perspectives =-=[10, 11, 12, 13, 14, 15, 16]-=-.s2.2. Relations Between Cohesion and SPssThe original definitions of a few of the measures investigated here consider whether tosconsider or ignore SPs in cohesion measurement. Bieman and Kang [4] su...

Citation Context

...echniques, and investigating the impact of this problem onsanother application of interest, predicting class instantiation reuse-proneness.s5s2.3. Relation Between Cohesion and Reuse-PronenesssBieman =-=[19]-=- identified two key class reuse mechanisms: instantiation (i.e., creating ansobject of a class) and inheritance. Each of these reuse mechanisms allows for the reuse ofsall members of a class except fo...

Citation Context

...ers of distinct attributes that are accessed by each method in a class.sCoh [3] Coh=a/kl, where a, k, and l have the same definitions as above.sLow-level designsSimilarity-based ClasssCohesion (LSCC) =-=[41]-=-s otherwise.s)1( )1( 1,ors0) and 0( ifs1 1, and 0 ifs0 )( 1 klk xx kkl kl CLSCC l i ii ,swhere l is the number of attributes, k is the number of methods, and xi is the nu...

Citation Context

...( ifs1 1, and 0 ifs0 )( 1 klk xx kkl kl CLSCC l i ii ,swhere l is the number of attributes, k is the number of methods, and xi is the number of methods thatsreference attribute i.sClass Cohesion (CC) =-=[42]-=- CC= The ratio of the sum of the similarities between all pairs of methods to the total number of pairs ofsmethods. The similarity between methods i and j is defined assji ji II II jiSimilarity )...

Citation Context

...imilar empirical studies, which allows us to compare our results with thosespreviously obtained. The selected cohesion measures have been well studied from bothstheoretical and empirical perspectives =-=[10, 11, 12, 13, 14, 15, 16]-=-.s2.2. Relations Between Cohesion and SPssThe original definitions of a few of the measures investigated here consider whether tosconsider or ignore SPs in cohesion measurement. Bieman and Kang [4] su...

Citation Context

...80% ≤ ROC < 90% indicates that the classification is excellent, and 90% ≤sROC ≤ 100% indicates that the classification is outstanding. Practical models are thosesthat have an ROC area of at least 70% =-=[32]-=-.sTo obtain more realistic and validated models, we applied the 10-cross-validationstechnique when constructing each univariate prediction model. For each model, wespartitioned the considered data set...

Citation Context

...ribute using asprediction model that incorporates the same cohesion measure.sSeveral researchers have qualitatively addressed the problem of considering or ignoringsSPs in cohesion measurement (e.g., =-=[3, 4, 5, 6, 7]-=-). To our knowledge, only Al Dallal [8]shas empirically studied the effect of considering or ignoring SPs. In that study, a group ofsexisting measures was selected and SPs were empirically investigate...

Citation Context

...imilar empirical studies, which allows us to compare our results with thosespreviously obtained. The selected cohesion measures have been well studied from bothstheoretical and empirical perspectives =-=[10, 11, 12, 13, 14, 15, 16]-=-.s2.2. Relations Between Cohesion and SPssThe original definitions of a few of the measures investigated here consider whether tosconsider or ignore SPs in cohesion measurement. Bieman and Kang [4] su...

Citation Context

...Each of these reuse mechanisms allows for the reuse ofsall members of a class except for private members. Class reuse-proneness is a key classsquality attribute of interest for software practitioners =-=[1]-=-. Several researchers (e.g., [4, 20,s21]) have explored the relationship between class reuse-proneness as an external qualitysattribute and several internal class quality attributes, such as cohesion,...

Citation Context

...imilar empirical studies, which allows us to compare our results with thosespreviously obtained. The selected cohesion measures have been well studied from bothstheoretical and empirical perspectives =-=[10, 11, 12, 13, 14, 15, 16]-=-.s2.2. Relations Between Cohesion and SPssThe original definitions of a few of the measures investigated here consider whether tosconsider or ignore SPs in cohesion measurement. Bieman and Kang [4] su...

Citation Context

...he same cohesion measure.sSeveral researchers have qualitatively addressed the problem of considering or ignoringsSPs in cohesion measurement (e.g., [3, 4, 5, 6, 7]). To our knowledge, only Al Dallal =-=[8]-=-shas empirically studied the effect of considering or ignoring SPs. In that study, a group ofsexisting measures was selected and SPs were empirically investigated to determine theirsimpact on the abil...

Citation Context

...imilar empirical studies, which allows us to compare our results with thosespreviously obtained. The selected cohesion measures have been well studied from bothstheoretical and empirical perspectives =-=[10, 11, 12, 13, 14, 15, 16]-=-.s2.2. Relations Between Cohesion and SPssThe original definitions of a few of the measures investigated here consider whether tosconsider or ignore SPs in cohesion measurement. Bieman and Kang [4] su...

Citation Context

...ip between class cohesion and instantiation reuseproneness. Barnard [20] empirical study results suggested that there is no relationshipsbetween LCOM1 and class reuse-proneness.sAl Dallal and Morasca =-=[22]-=- empirically investigated the relations between 21 classscohesion measures and both inheritance and instantiation reuse-proneness. They proposedsstatistical models to predict classes prone to be (1) r...

Citation Context

... scenario to another as a critically predicted class (CPC). Thesexistence of CPCs provides evidence that the type of SPs considered can alter thesprediction results.sUsing Shapiro-Wilk normality test =-=[33]-=-, we evaluated the distribution normality of thesprobability estimations obtained using each of the prediction models built for eachsconsidered measures in each scenario. None of them found to follow ...

Citation Context

...ned. The ICBMC and OL2 measures were originally defined tosevaluate cohesion when ignoring SPs. Because constructors and destructors do notsalways access all attributes in a class, Fernandez and Pena =-=[18]-=- suggested consideringsconstructors and destructors in cohesion measurement. In addition, they argued thatsaccess methods must be ignored because they always access a single attribute.sAl Dallal [8] e...

Citation Context

...tysWe only accounted for Java systems, which disallowed ussfrom empirically exploringsthe impact of considering destructors. However, constructors and destructors are expectedsto have similar effects =-=[8, 35]-=-. In this paper, we considered two well-known and widelysused systems within the research community; Eclipse and JHotDraw. Although thesessystems are open-source systems that may not be representative...

Citation Context

...nitions of the considered cohesion measuressTable A.1: Definitions of the class cohesion measures under considerationsClass CohesionsMeasuresDefinition/FormulasThe lack of Cohesion insMethods (LCOM1) =-=[36]-=-sLCOM1= Number of pairs of methods that do not share attributes.sLCOM2 [37] P= Number of pairs of methods that do not share attributes.sQ= Number of pairs of methods that share attributes.s ...

Citation Context

...ribute using asprediction model that incorporates the same cohesion measure.sSeveral researchers have qualitatively addressed the problem of considering or ignoringsSPs in cohesion measurement (e.g., =-=[3, 4, 5, 6, 7]-=-). To our knowledge, only Al Dallal [8]shas empirically studied the effect of considering or ignoring SPs. In that study, a group ofsexisting measures was selected and SPs were empirically investigate...

Citation Context

...llows for the reuse ofsall members of a class except for private members. Class reuse-proneness is a key classsquality attribute of interest for software practitioners [1]. Several researchers (e.g., =-=[4, 20,s21]-=-) have explored the relationship between class reuse-proneness as an external qualitysattribute and several internal class quality attributes, such as cohesion, coupling, and size.sThey theoretically ...

Citation Context

...llection ProcesssWe applied logistic regression (LR) [25], a statistical technique that is widely used tosbuild prediction models for external quality attributes, such as class fault-proneness (e.g.,s=-=[3, 16, 26, 27]-=-), class reuse-proneness (e.g., [22]), and class maintainability (e.g., [28]).sApplying other statistical techniques to construct prediction models, such as those usedsby Briand and Wuest [29], Subram...

Citation Context

... is widely used tosbuild prediction models for external quality attributes, such as class fault-proneness (e.g.,s[3, 16, 26, 27]), class reuse-proneness (e.g., [22]), and class maintainability (e.g., =-=[28]-=-).sApplying other statistical techniques to construct prediction models, such as those usedsby Briand and Wuest [29], Subramanyam and Krishnan [30], and Arisholm et al. [31],sand comparing the results...

Citation Context

...eness (e.g., [22]), and class maintainability (e.g., [28]).sApplying other statistical techniques to construct prediction models, such as those usedsby Briand and Wuest [29], Subramanyam and Krishnan =-=[30]-=-, and Arisholm et al. [31],sand comparing the results with the LR results are outside the scope of this paper.sIn LR analysis, the variables are classified as dependent or independent. The independent...

Citation Context

...that do not share attributes.sLCOM2 [37] P= Number of pairs of methods that do not share attributes.sQ= Number of pairs of methods that share attributes.s otherwise QPifQP LCOM 0 0 2sLCOM3 =-=[38]-=- LCOM3= Number of connected components in the graph that represents each method as a node and thessharing of at least one attribute as an edge.sLCOM4 [39] Similar to LCOM3, where additional edges are ...