Cross versus Within-Company Cost Estimation Studies: A Systematic Review

Kitchenham, Barbara; Mendes, Emilia; Travassos, Guilherme Horta

doi:10.1109/tse.2007.1001

Cited by 321 publications

(251 citation statements)

References 16 publications

Supporting

Mentioning

244

Contrasting

Unclassified

Order By: Relevance

“…In all the other cases, where the actual height lies somewhere within the range of predictions, the median will be more accurate than at least 50% of the individual predictions. If, for example, the actual height were 125 metres, the median of 26.5 metres would be better than seven of the predictions (10,15,15,20,21,23, and 370 metres).…”

Section: Combinations Of Time Predictionsmentioning

confidence: 88%

“…When developing a formal model to predict time usage, it is usually safest to use local data, such as data from your own context, the organization where you work, or at least data from very similar contexts [20]. Time prediction models intended to be applicable across a range of contexts are typically less accurate [19].…”

Section: Time Prediction Modelsmentioning

confidence: 99%

“…Another group of students used the Fibonacci-inspired, nonlinear scale 1, 2, 3,5,8,13,20,30,40, which has the middle number 8 [50]. The students were asked to predict the number of work hours they would use for a software development task using numbers from the scale they had been assigned.…”

Section: From Fibonacci To T-shirt Sizes: Time Predictions Using Altementioning

confidence: 99%

See 2 more Smart Citations

Time Prediction Methods and Principles

Halkjelsvik

Jørgensen

2018

Time Predictions

View full text Add to dashboard Cite

The prominent approach for reducing a problem's complexity is to decompose it into less complex subproblems, solve each of these, and then aggregate the subsolutions into an overall solution. In time prediction contexts, this approach is typically the basis of what has been referred to as the bottom-up method, the activity-based method, or predictions based on a work breakdown structure. Generally, across a range of domains, decomposition has been found to improve judgement quality and increase prediction accuracy [1]. In the domain of time predictions, however, there are also situations in which decomposition leads to overoptimistic and less accurate judgements [2].The time prediction literature has examined two types of decomposition strategies: unpacking, which consists of merely listing or thinking about the subcomponents of a task before predicting the time usage as a whole, and decomposition, which consists of unpacking, predicting the time usage for each unpacked component, and then aggregating the time predictions into a prediction of the total time usage.Unpacking: Unpacking tend to give higher time predictions. For example, if you ask people to list all the persons for whom they must buy Christmas presents, they will tend to predict that they need more time to complete their Christmas shopping compared to those who did not generate such a list. Unpacking strategies may be based either on the self-generation of components, as in the example above, or on reminding the participants of possible subcomponents, for example, by using a checklist for activities to be included. Checklists, or reminders, are usually very effective and easy to implement and may improve the accuracy of time predictions, particularly in situations in which there is a tendency towards overoptimistic time predictions [3]. In one study illustrating how unpacking tends to increase time predictions, participants were instructed to format a document to match a printed, edited version of the same document. When asked to predict how long it would take to format the document without any reminders of the work's components, the participants predicted, on average, eight and a half minutes. When asked to predict how long it would take with reminders of the work's components, such as including italics and special characters (ђ,ǒ, î), the participants predicted, on average, about 13 minutes [4].In many cases, the increase in time predictions from unpacking a task contributes to greater realism [5]. Projects that used checklists when predicting time were, for example, more accurate and less overoptimistic (with predictions, on average, 10% too low) than projects that did not (with predictions, on average, 38% too low) [3]. Although unpacking may generally contribute to more accurate and less overoptimistic time predictions, this may not always be the case. Pointing out obvious components of a task, small and simple components, or components that are part of the last steps of a task may not lead to more accurate time predictions [4].There are also ...

show abstract

Section: Combinations Of Time Predictionsmentioning

confidence: 88%

Section: Time Prediction Modelsmentioning

confidence: 99%

See 1 more Smart Citation

Time Prediction Methods and Principles

Halkjelsvik

Jørgensen

2018

Time Predictions

View full text Add to dashboard Cite

show abstract

“…The search process describes the process to identify the list of candidate studies. Following search process advocated by Barbara Kitchenham et al [7], the search process was organized into two separate phases. The initial search phase identified candidate primary studies based on searches of electronic digital libraries from IEEE, ACM, Elsevier, Springer, and Wiley.…”

Section: Rq3 How Can External Validation Of the Prediction Model Be mentioning

confidence: 99%

A Framework for Defect Prediction in Specific Software Project Contexts

Wahyudin

Ramler

Biffl

2011

Software Engineering Techniques

View full text Add to dashboard Cite

Abstract. Software defect prediction has drawn the attention of many researchers in empirical software engineering and software maintenance due to its importance in providing quality estimates and to identify the needs for improvement from project management perspective. However, most defect prediction studies seem valid primarily in a particular context and little concern is given on how to find out which prediction model is well suited for a given project context. In this paper we present a framework for conducting software defect prediction as aid for the project manager in the context of a particular project or organization. The framework has been aligned with practitioners' requirements and is supported by our findings from a systematical literature review on software defect prediction. We provide a guide to the body of existing studies on defect prediction by mapping the results of the systematic literature review to the framework.

show abstract

“…However, companies may face difficulties in terms of WC data collection, e.g. Kitchenham et al (2007), (1) the time required to accumulate enough data (Cherkassky and Mulier 1998) on past projects from a single company may be prohibitive; (2) by the time the dataset is large enough (Cherkassky and Mulier 1998), technologies used by the company may have changed, and older projects may no longer be representative of current practices; and (3) care is necessary as data need to be collected in a consistent manner.…”

Section: Introductionmentioning

confidence: 99%

Which models of the past are relevant to the present? A software effort estimation approach to exploiting useful past models

Minku

Yao

2016

Autom Softw Eng

View full text Add to dashboard Cite

Software Effort Estimation (SEE) models can be used for decision-support by software managers to determine the effort required to develop a software project. They are created based on data describing projects completed in the past. Such data could include past projects from within the company that we are interested in (WC projects) and/or from other companies (cross-company, i.e., CC projects). In particular, the use of CC data has been investigated in an attempt to overcome limitations caused by the typically small size of WC datasets. However, software companies operate in non-stationary environments, where changes may affect the typical effort required to develop software projects. Our previous work showed that both WC and CC models of the past can become more or less useful over time, i.e., they can sometimes be helpful and sometimes misleading. So, how can we know if and when a model created based on past data represents well the current projects being estimated? We propose an approach called Dynamic Cross-company Learning (DCL) to dynamically identify which WC or CC past models are most useful for making predictions to a given company at the present. DCL automatically emphasizes the predictions given by these models in order to improve predictive performance. Our experiments comparing DCL against existing WC and CC approaches show that DCL is successful in improving SEE by A preliminary version of this work appeared as Minku and Yao (2012a

show abstract

Cross versus Within-Company Cost Estimation Studies: A Systematic Review

Cited by 321 publications

References 16 publications

Time Prediction Methods and Principles

Time Prediction Methods and Principles

A Framework for Defect Prediction in Specific Software Project Contexts

Which models of the past are relevant to the present? A software effort estimation approach to exploiting useful past models

Contact Info

Product

Resources

About