ABSTRACT Software development effort estimation (SDEE) is a key task in managing software projects. Among the existing SDEE models, artificial neural networks (ANN) have garnered considerable attention from the software engineering community because of their ability to learn from previous data and yield acceptable estimates. However, to the best of the authors' knowledge, no systematic literature review (SLR) has been carried out with focus on the use of ANNs in SDEE. This work aims to analyze ANN‐based SDEE studies from five view‐points: estimation accuracy, accuracy comparison, estimation context, impact of combining ANN‐based SDEE models with other techniques, and ANNs parameters. To find relevant ANN‐based SDEE studies, we carried out an automated search using four electronic databases. The quality of the relevant papers was assessed to determine the set of papers to include in our review. We identified 65 papers published in the period 1993–2023 with acceptable quality score. The results of our systematic review revealed that ANN‐based SDEE models perform better than 11 machine learning (ML) and non‐ML SDEE models. Further, the estimation accuracy is improved when neural networks are used in combination with other techniques such as fuzzy clustering techniques. This study found that the use of ANN models in SDEE is promising to get accurate estimates. However, the application of ANN models in industry is still limited. Therefore, it is recommended that practitioners cooperate with researchers to encourage and facilitate the application of ANN models in industry.

Abstract During the past decades, many studies have been carried out in an attempt to build accurate software development effort estimation techniques. However, none of the techniques proposed has proven to be successful at predicting software effort in all circumstances. Among these techniques, analogy‐based estimation has gained significant popularity within software engineering community because of its outstanding performance and ability to mimic the human problem solving approach. One of the challenges facing analogy‐based effort estimation is how to predict effort when software projects are described by a mixture of continuous and categorical features. To address this issue, the present study proposes an improvement of our former 2FA‐kprototypes technique referred to as 2FA‐cmeans. 2FA‐cmeans uses a clustering technique, called general fuzzy c‐means, which is a generalization of the fuzzy c‐means clustering technique to cluster objects with mixed attributes. The performance of 2FA‐cmeans was evaluated and compared with that of our former 2FA‐kprototypes technique as well as classical analogy over six datasets that are quite diverse and have different sizes. Empirical results showed that 2FA‐cmeans outperforms the two other analogy techniques using both all‐in and jackknife evaluation methods. This was also confirmed by the win–tie–loss statistics and the Scott–Knott test.

Producing reliable and accurate estimates of software effort remains a difficult task in software project management, especially at the early stages of the software life cycle where the information available is more categorical than numerical.In this paper, we conducted a systematic mapping study of papers dealing with categorical data in software development effort estimation.In total, 27 papers were identified from 1997 to January 2019.The selected studies were analyzed and classified according to eight criteria: publication channels, year of publication, research approach, contribution type, SDEE technique, Technique used to handle categorical data, types of categorical data and datasets used.The results showed that most of the selected papers investigate the use of both nominal and ordinal data.Furthermore, Euclidean distance, fuzzy logic, and fuzzy clustering techniques were the most used techniques to handle categorical data using analogy.Using regression, most papers employed ANOVA and combination of categories.

Abstract Analogy‐based estimation is one of the most widely used techniques for effort prediction in software engineering. However, existing analogy‐based techniques suffer from an inability to correctly handle nonquantitative data. To deal with this limitation, a new technique called 2FA‐kprototypes was proposed and evaluated. 2FA‐kprototypes is based on the use of the fuzzy k‐prototypes clustering technique. Although fuzzy k‐prototypes algorithms are well known for their efficiency in clustering numerical and categorical data, they are sensitive to the selection of initial cluster centers. In this paper, the impact of cluster center initialization on improving the prediction accuracy of 2FA‐kprototypes was analyzed and discussed using two cluster initialization techniques: centrality‐based initialization and density‐based initialization. The performance of 2FA‐kprototypes using these two initialization techniques was evaluated and compared with that of 2FA‐kprototypes using random initialization over four datasets: ISBSG, COCOMO81, USP05‐FT, and USP05‐RQ. The results showed an improvement in the performance of 2FA‐kprototypes in terms of estimation accuracy when the all‐in method is used.

Estimation by analogy is a commonly used software effort estimation technique and a suitable alternative to other conventional estimation techniques: It predicts the effort of the target project using information from former similar projects. While it is relatively easy to handle numerical attributes, dealing with categorical attributes is one of the most difficult issues for analogy-based estimation techniques. Therefore, we propose, in this paper, a novel analogy-based approach, called 2FA-kprototypes, to predict effort when software projects are described by a mix of numerical and categorical attributes. To this aim, the well-known fuzzy k-prototypes algorithm is integrated into the process of estimation by analogy. The estimation accuracy of 2FA-kprototypes was evaluated and compared with that of two techniques: (1) classical analogy-based technique and (2) 2FA-kmodes, which is a technique that we have developed recently. The comparison was performed using four data sets that are quite diverse and have different sizes: ISBSG, COCOMO, USP05-FT, and USP05-RQ. The results obtained showed that both 2FA-kprototypes and 2FA-kmodes perform better than classical analogy.

Analogy-based estimation has recently emerged as a promising technique and a viable alternative to other conventional estimation methods. One of the most important research areas for analogy-based cost estimation is how to predict the effort of software projects when they are described by mixed numerical and categorical data. To address this issue, we have proposed, in an earlier work, a new approach called fuzzy analogy combining the key features of fuzzy logic and analogy-based reasoning. However, fuzzy analogy may only be used when the possible values of the categorical attributes are derived from a numerical domain. The current study aims to extend our former approach to correctly handle categorical data. To this end, the fuzzy k-modes algorithm is used with two initialization techniques. The performance of the proposed approach was compared with that of classical analogy using the International Software Benchmarking Standards Group (ISBSG) dataset. The obtained results show significant improvement in estimation accuracy.

Analogy-based software development effort estimation methods have proved to be a viable alternative to other conventional estimation methods since they mimic the human problem solving approach. However, they are limited by their inability to correctly handle categorical data. Therefore, we have proposed, in an earlier work, a new approach called fuzzy analogy which extends classical analogy by incorporating the fuzzy logic concept in the estimation process. The proposed approach may be applied only when the categorical values are derived from numerical data. This paper extends fuzzy analogy to deal with categorical values that are not derived from numerical data. To this aim, we used the fuzzy k-modes algorithm, a well-known clustering technique for large datasets containing categorical values. Thereafter, we evaluate the accuracy of fuzzy analogy construction-based on fuzzy k-modes using the ISBSG R8 dataset. This evaluation shows that our proposed approach leads to significant improvement in estimation accuracy.

Software effort estimation is one of the most important tasks in software project management. Of several techniques suggested for estimating software development effort, the analogy-based reasoning, or Case-Based Reasoning (CBR), approaches stand out as promising techniques. In this paper, the benefits of using linguistic rather than numerical values in the analogy process for software effort estimation are investigated. The performance, in terms of accuracy and tolerance of imprecision, of two analogy-based software effort estimation models (Classical Analogy and Fuzzy Analogy, which use numerical and linguistic values respectively to describe software projects) is compared. Three research questions related to the performance of these two models are discussed and answered. This study uses the International Software Benchmarking Standards Group (ISBSG) dataset and confirms the usefulness of using linguistic instead of numerical values in analogy-based software effort estimation models.