Spectral Clustering Effect in Software Development Effort Estimation

Software development effort estimation is essential for software project planning and management. In this study, we present a spectral clustering algorithm based on symmetric matrixes as an option for data processing. It is expected that constructing an estimation model on more similar data can increase the estimation accuracy. The research methods employ symmetrical data processing and experimentation. Four experimental models based on function point analysis, stepwise regression, spectral clustering, and categorical variables have been conducted. The results indicate that the most advantageous variant is a combination of stepwise regression and spectral clustering. The proposed method provides the most accurate estimates compared to the baseline method and other tested variants.

Investigating the Use of an Adaptive Neuro-Fuzzy Inference System in Software Development Effort Estimation

Modeling software development effort estimation models has been a hot research topic over the last three decades. Numerous models were proposed in these decades to predict the effort. The key challenges for future software development is providing accurate software estimation. Failure to acknowledge the accuracy of effort estimation can cause inaccurate estimation, customer disappointment, and poor software development or project failure. This research presents a novel computational technique, named adaptive neuro-fuzzy inference system (ANFIS), for the modeling of software effort estimation. It was developed utilizing the Constructive Cost Model (COCOMO) dataset. The data were randomly divided into two sets: 83% for training and 17% for testing. The mean magnitude relative-error (MMRE) and the coefficient of correlation (R) were used as assessment indices. Results showed that the accuracy of the proposed model is quite satisfactory in comparison with actual values. Moreover, a comparison study was conducted with another approach. The results showed that ANFIS produced better results in comparison with Function Point Analysis, Software Lifecycle Management, and COCOMO methods. ANFIS was found to be a potential predictive model for software development effort estimation.