Understanding Project Data

The objective of this chapter is to provide an introduction to statistical techniques and concepts that are frequently used when dealing with data for effort estimation. The concepts presented here are in no way exhaustive since statistics comprises a very large body of knowledge where entire books are devoted to specific topics. The parts that are the focus of this chapter are those that are necessary to use when building effort estimation models, and also when comparing different effort estimation techniques.

What is the Best Technique?

Although numerous studies on Web effort estimation have been carried out to date, there is no consensus on what constitutes the best effort estimation technique to be used by Web companies. It seems that not only the effort estimation technique itself can influence the accuracy of predictions, but also the characteristics of the data set used (e.g., skewness, collinearity; Shepperd & Kadoda, 2001). Therefore, it is often necessary to compare different effort estimation techniques, looking for those that provide the best estimation accuracy for the data set being employed. With this in mind, the use of graphical aids such as boxplots is not always enough to assess the existence of significant differences between effort prediction models. The same applies to measures of prediction accuracy such as the mean magnitude of relative error (MMRE), median magnitude of relative error (MdMRE), and prediction at level l (Pred[25]). Other techniques, which correspond to the group of statistical significance tests, need to be employed to check if the different residuals obtained for each of the effort estimation techniques compared come from the same population. This chapter details how to use such techniques and how their results should be interpreted.

How to Improve Your Company's Effort Estimation Practices

Numerous Web development companies worldwide do not employ formal techniques to estimate effort for new projects, thus relying on expert-based opinion (McDonald & Welland, 2001; Mendes, Mosley, & Counsell, 2005). In addition, many Web companies do not gather any data on past projects, which can later be used to estimate effort for new projects, and as a consequence they are not aware of how effort is used throughout their projects and if it could be used more effectively. This chapter provides a set of guidelines we believe can be of benefit to Web companies to help them improve their effort estimation practices. Our guidelines are particularly targeted at small Web development companies.

Web Effort Estimation Using Case-Based Reasoning

Software practitioners recognise the importance of realistic effort estimates to the successful management of software projects, the Web being no exception. Having realistic estimates at an early stage in a project’s life cycle allow project managers and development organisations to manage resources effectively. Several techniques have been proposed to date to help organisations estimate effort for new projects. One of these is a machine-learning technique called case-based reasoning. This chapter presents a case study that details step by step, using real data from completed industrial Web projects, how to obtain effort estimates using case-based reasoning, and how to assess the prediction accuracy of this technique. The reason to describe the use of case-based reasoning for effort estimation is motivated by its previous use with promising results in Web effort estimation studies.

What is Different about Web and Software Projects?

The objective of this chapter is threefold. First is to introduce new terminology that relates specifically to hypertext, the model the Web is based upon. Second, it provides an overview of differences between Web and software development with respect to their development processes, technologies, quality factors, and measures. Third, it discusses the differences between Web effort estimation and software effort estimation.

How Accurate is an Effort Model?

Building effort models or using techniques to obtain a measure of estimated effort does not mean that the effort estimates obtained will be accurate. As such, it is also important and necessary to assess the estimation accuracy of the effort models or techniques under scrutiny. For this, we need to employ a process called cross-validation. Cross-validation means that part of the original data set is used to build an effort model, or is used by an effort estimation technique, leaving the remainder of the data set (data not used in the model-building process) to be used to validate the model or technique. In addition, in parallel with conducting cross-validation, prediction accuracy measures are also obtained. Examples of de facto accuracy measures are the mean magnitude of relative error (MMRE), the median magnitude of relative error (MdMRE), and prediction at 25% (Pred[25]).

Introduction to Web Cost Estimation

Software effort models and effort estimates help project managers allocate resources, control costs, and schedule and improve current practices, which in theory should allow projects to be finished on time and within budget. In the context of Web development and maintenance, these issues are also crucial and very challenging given that Web projects have short schedules and a highly fluidic scope. Therefore, the objective of this chapter is to introduce the concepts related to Web effort estimation and effort estimation techniques. These concepts will be used in further chapters.

The Need for Empirical Web Engineering

The objective of this chapter is to motivate the need for empirical investigations in Web engineering, and additionally to describe the three main types of empirical investigations that can be used by Web companies to understand, control, and improve the products they develop and the processes they use. These three main types of empirical investigations are surveys, case studies, and formal experiments. Although all these three types are described in this chapter, we focused our attention on formal experiments as these are the most difficult type of investigation to plan and execute.

Sizing Web Applications

Surveying and classifying previous work in a particular field brings several benefits, which are to (a) help organise a given body of knowledge, (b) provide results that can help identify gaps that need to be filled, (c) provide a categorisation that can also be applied or adapted to other surveys, and (d) provide a classification and summary of results that may benefit practitioners and researchers who wish to carry out meta-analyses. This chapter presents a literature survey of size measures (attributes) that have been published since 1992 and classifies these measures according to a proposed taxonomy. We also discuss ways in which Web companies can devise their own size measures.

Web Effort Estimation Using Classification and Regression Trees

The use of realistic effort estimates is fundamental to both software and Web project management as they help project managers allocate resources, control costs and schedule, and improve current practices, leading to projects that are finished on time and within budget. Different effort techniques have been used to obtain effort estimates for Web projects. Two—stepwise regression and case-based reasoning—have already been presented in Chapters V and VI respectively. In this chapter we detail a third technique used to obtain effort estimates for Web projects, known as classification and regression trees (CART), that is considered a machine-learning technique. We detail its use by means of a case study where a real effort prediction model based on data from completed industrial Web projects is constructed step by step.