ENHANCED EFFORT ESTIMATION BY EXTENDING BASIC PROGRAMMING MODELS TO INCLUDE MODULARITY FACTORS
Abstract
Because of the labor intensive nature of software, there is an urgent need to understand and measure the underlying causes of complexity so that software costs may be reduced. To increase understanding, we have developed models of modularity factors which have been used to better comprehend both program construction and comprehension. The two main models developed are the logical module model and the interconnection model. The first reflects the logical segmentation of large physical modules. The second model enables the estimation of complexity associated with module coupling. The models were validated using two sets of experimental data which were collected on two different occasions. Further support was given when the models were able to explain the results of a comprehension experiment, which was conducted to determine how comment and modularization factors affect comprehension. The extended complexity measure was able to account for 80 percent of the variance in the data's complexity. This was an improvement of 30 percent over the original effort measure. The inclusion of all factors reduced the average relative error from 150 percent to 1 percent. The logical module model enabled us to understand why the monolithic program from the comprehension experiment was easily understood. The interconnection model showed the perfect structure of the abstract data type modularization version and indicated why module connections can cause functionally modularized programs to be complex. Although the programs used as data were quite small, we feel that the basic factors affecting these programs need to be understood and measured. Only then can we begin to understand the other factors affecting larger programs.
Degree
Ph.D.
Subject Area
Computer science
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.