Mikael Lindvall and Kristian Sandahl. How well do experienced software developers predict software change? The Journal of Systems and Software, 43(1):19--27, 1998.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....quite stable in the behavior it is supposed to provide. The mediation domain is considerable less stable and the developers of the framework have had to incorporated impressive amounts of requirement changes and additions during its life time. 6 RELATED WORK Recent work by Lindvall and Sandahl [19] shows that software developers are not so good at predicting from a requirements specification how many and which classes that will be changed. Their empirical study shows that only between 30 40 of the actual changed classes were predicted to be changed by the developers. This can be seen as an ....

Lindvall, M. and Sandahl, K. (1998). How Well do Experienced Software Developers Predict Software Change? Journal of Systems and Software, 43(1):19 -- 27.

....design of the framework and information about where likely changes occur may reduce the cost of redesign the framework to meet the new demands. Studies show that software engineers can only identify a subset ( 50 percent) of future changes and they can not provide the complete picture of change [7]. Thus, a more objective method for identifying structural shortcomings will provide more reliable input for maintenance effort estimations and thereby more accurate estimates. Recently an approach for identifying potential modules for restructuring in large software systems (about 10 millions ....

....of a large telecommunication switching system. Our work uses a similar approach but on a smaller system, 300 600 classes, and we use historical information about modules and classes, which are entities of smaller granularity than in the work Gall et al. Recent work by Lindvall and Sandahl [7] shows that software developers are not so good at predicting from a requirements specification how many and which classes that will be changed. Their empirical study shows that only between 30 40 percent of the actual changed classes were predicted to be changed by the developers. This can be ....

Lindvall, M. and Sandahl, K. (1998). How Well do Experienced Software Developers Predict Software Change? Journal of Systems and Software, 43(1):19 -- 27.

....for restructuring. The information about where likely changes occur may reduce the cost of redesign the framework. Studies show that software engineers without automated support can only identify a subset ( 50 percent) of future changes and they can not provide the complete view of change [15]. Thus, a more objective method for identifying structural shortcomings in an object oriented framework will provide more reliable input for maintenance effort estimations and thereby increase the accuracy. Framework Deployment: Assessment of the framework to decide when it can be released for ....

....switching system. Our work uses a similar approach but on a smaller and object oriented system, 300 600 classes, and we use historical information about modules and classes, which are entities of smaller granularity than in the work Gall et al. Recent work by Lindvall and Sandahl [15] shows that software developers are not so good at predicting from a requirements specification how many and which classes that will be changed. Their empirical study shows that only between 30 40 percent of the classes actually changed were predicted to be changed by the developers. This can be ....

Lindvall, M. and Sandahl, K. (1998). How Well do Experienced Software Developers Predict Software Change? Journal of Systems and Software, 43(1):19 -- 27.

....documents, and a graph theory model. Other authors [7, 17] have discussed impact analysis methods focused on source code. Based on the source code and a particular change to that source code, Kung et al. 17] propose a way to do automatic ripple effects analysis of the change. Lindvall and Sandahl [24] have reported on a study on the accuracy of the designers in predicting necessary changes. Empirical results on source code impact analysis indicate that software engineers, when doing impact analysis predict only half of the necessary changes. However, it is noted that the modifications ....

M. Lindvall and K. Sandahl, `How Well do Experienced Software Developers Predict Software Change?', Journal of Systems and Software, 43(1): 19-27, 1998.

....post mortem analysis of the change process in this perspective was made available to project developers and project management. 4. BACKGROUND Our previous analyses were concerned with evaluation ofRDIA as conducted on the class level over one release (R4) 7] and over two releases (R4 and R6) [5]. We also analysed RDIA on the individual requirements level [6] over one release (R4) The results from these analyses were that the correctness of the prediction is high (i.e. classes predicted to change do change) while completeness is low (i.e. many more classes than predicted were actually ....

....than lines of code, and classes, which is more coarse grained than member functions. In this study, we compare the prediction mainly in terms of the number of predicted and changed entities. Correctness and completeness of the prediction on the class level were calculated in previous studies (e.g. [5]) but required more details than were available on the member function level. The number of member functions is still important to consider as the number of predicted member functions was used as a basis for estimating the effort (the cost) required in the PMRproject. The effort was calculated by ....

Lindvall, M. and Sandahl, K. 1998. How Well do Experienced Software Developers Predict Software Change? The Journal of Systems and Software, 43(1):19 -- 27. 16

.... requirements, as well as the code base (unless the design was made with such changes in mind, explicitly) In our cxperiencc requirements engineers are fairly capable of pointing out requirements that will have a large impact of other parts of a system, as part of the cost estination procedures [7]. Thus we believe that identifica tion of the 20 most dependent requirements could be a surmountable problem, especially if supported by heuristics of the kind listed above. 3.5. An interdependency measure Sclccting a number of requirements for realization in a particular release can be viewed ....

Lindvail, M. and Sandahl, K. How Well do Experienced Software Developers Predict Software Change? Journal of Sys- tems and Software, 43(1 ): 19 - 27, 1998.

....factor (the number of changed entities divided by the number of predicted entities) varied between 1.5 and 3.1 on the class level and between 4.0 and 6. 1 4 on the member function level [5] We also found that developers were unaware of their own capabilities in predicting change [9] [6] and that developers were reluctant to using object models actively during maintenance [10] This work was motivated by the large problem of underprediction and the implied need to build knowledge about what characterizes changed classes. If, for example, it is a general pattern that large ....

Lindvall, M. and Sandahl, K. How Well do Experienced Software Developers Predict Software Change? The Journal of Systems and Software, 1997. Accepted for publication.

No context found.

M. Lindvall and K. Sandahl, How Well do Experienced Software Developers Predict Software Change? Journal of Systems and Software, 43(1):19-27, 1998.

No context found.

Mikael Lindvall and Kristian Sandahl. How well do experienced software developers predict software change? The Journal of Systems and Software, 43(1):19--27, 1998.

No context found.

M. Lindvall and K. Sandahl. How well do experienced software developers predict software change? Journal of Systems and Software, 43(1):19--27, Oct. 1998.

No context found.

M. Lindvall and K. Sandahl. How well do experienced software developers predict software change? Journal of Systems and Software, 43(1):19--27, Oct. 1998.

No context found.

M. Lindvall and K. Sandahl. How well do experienced software developers predict software change? Journal of Systems and Software, 43(1):19--27, Oct. 1998.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC