Use Case Point Approach Based Software Effort Estimation using Various Support Vector Regression Kernel Methods

The job of software effort estimation is a critical one in the early stages of the software development life cycle when the details of requirements are usually not clearly identified. Various optimization techniques help in improving the accuracy of effort estimation. The Support Vector Regression (SVR) is one of several different soft-computing techniques that help in getting optimal estimated values. The idea of SVR is based upon the computation of a linear regression function in a high dimensional feature space where the input data are mapped via a nonlinear function. Further, the SVR kernel methods can be applied in transforming the input data and then based on these transformations, an optimal boundary between the possible outputs can be obtained. The main objective of the research work carried out in this paper is to estimate the software effort using use case point approach. The use case point approach relies on the use case diagram to estimate the size and effort of software projects. Then, an attempt has been made to optimize the results obtained from use case point analysis using various SVR kernel methods to achieve better prediction accuracy.

💡 Research Summary

The paper addresses the long‑standing challenge of accurately estimating software development effort during the early phases of a project, when detailed requirements are often unavailable. Traditional size‑based estimation techniques such as Lines of Code (LOC) and Function Points (FP) are tied to procedural paradigms and suffer from language‑dependence and subjectivity, limiting their usefulness for object‑oriented development. To overcome these drawbacks, the authors adopt the Use Case Point (UCP) model, originally proposed by Gustav Karner in 1993, which derives a size metric from the UML use‑case diagram. UCP calculation proceeds through four steps: (1) classification of actors and use cases into simple, average, or complex categories, each assigned a weighting factor; (2) aggregation of weighted actor and use‑case counts to obtain Unadjusted Actor Weight (UAW) and Unadjusted Use‑Case Weight (UUCW); (3) adjustment of the unadjusted total (UUCP) using Technical Complexity Factor (TCF) and Environmental Factor (EF) scores derived from 13 and 8 questionnaire items respectively; and (4) multiplication of UUCP by TCF and EF to produce the final UCP value. While UCP provides a systematic, diagram‑driven size estimate, it assumes a linear relationship between size and effort, an assumption that is rarely true in practice.

To capture the non‑linear relationship, the authors integrate Support Vector Regression (SVR), a kernel‑based machine‑learning technique that maps input data into a high‑dimensional feature space via a kernel function K(xi, xj) = φ(xi)·φ(xj) and then finds an ε‑insensitive hyperplane that minimizes a regularized risk function. Four kernels are examined: linear, polynomial (degree d, coefficient r), Radial Basis Function (RBF) with width parameter γ, and sigmoid (tanh) with parameters γ and r. The SVR formulation includes the penalty parameter C, slack variables ξ, ξ*, and the ε‑tube that defines the region of zero loss.

The methodology consists of the following pipeline: (i) compute UCP for each historical project; (ii) normalize UCP values to the