Review on Cost Estimation Techniques for Web Projects [Part 1]

Review on Cost Estimation Techniques for Web Projects

Prepared by Sayed Mohsin Reza

Part 1

Source Book: Cost Estimation Techniques for Web Projects

Book Author: Emilia Mendes University of Auckland, New Zealand

Table of Contents

• Book Objective/ Purpose:
• Difference about Web and Software Projects
• Introduction to Web Cost Estimation
• Accuracy of an Effort Model?
• Sizing Web Applications

Book Objective/ Purpose:

The objective of this book is therefore to provide Web companies, researchers, and students with the necessary knowledge on Web effort and cost estimation.

It includes step-by-step guidelines on how to use and compare several effort estimation techniques, which may considerably help companies improve their current effort estimation practices, and help researchers and students understand the process that needs to be carried out to estimate development effort.

Difference about Web and Software Projects

Web applications into three different categories

• Web hypermedia application: An application characterized by the authoring of information using nodes (chunks of information), links (relations between nodes), anchors, access structures (for navigation), and delivery over the Web
• Web software application: An application that often implements a conventional business domain and uses the Web’s infrastructure for execution
• Web application: combines characteristics of both Web hypermedia and Web software applications

Differences between Web and conventional software development

• Application Characteristics and Availability

·         determine web application or Traditional software

·         Web applications are distributed, are cross-platform, integrate numerous distinct components, and contain content that is structured using navigational structures with hyperlinks.

·         Traditional software applications are generally monolithic and single platform, and can integrate distinct components.

1. Technology and Architecture

·         Web application: Determine whether the application is built on Java solutions, HTML, JavaScript, XML, UML, databases, third-party components and middleware, and so forth. In terms of their architecture, two-tier or an n-tier

·         Traditional Software: object-oriented languages, relational databases, and CASE tools

c.       Quality Drivers

·         Web application: quality product, reliability, usability, and security.

·         Traditional Software: quality product

d.      Information Structuring, Design, and Maintenance

·         web application: structured/unstructured, hyperlinks

·         Traditional software: structured and seldom employ hyperlinks.

e.       Disciplines and People Involved in Development

·         Web application: software/ hypermedia/ requirements / usability /information engineering/ graphics design/ network management

·         Traditional Software: programming, database design, and project management

f.       Stakeholders

g.      Legal, Social, and Ethical Issues

Introduction to Web Cost Estimation

Effort estimation enables companies to know beforehand and before implementing an application the amount of effort required to develop the application on time and within budget.

Main goal is to understand the project variables that may affect effort prediction to estimate the web cost of a project.

Figure 1: Steps used to obtain an effort estimate

Several mechanisms were established to understand the project variables. Some of are –

·         Expert-based estimation

·         Algorithmic-based estimation

·         artificial-intelligence techniques

Expert Based estimation: process of estimating effort by subjective means. This estimation based on previous experience with developing and/or managing similar projects. Effort estimation is directly proportional to the competence and experience of the individuals.

Drawbacks of expert-based estimation

1.      Repeatability

2.      experience alone is not enough to identify the underlying relationship between effort and size-cost drivers

3.      Optimistic estimates lead to underestimated effort with the direct consequence of projects being over budget and late.

Algorithmic based estimation: most popular techniques in the Web and software effort estimation. It is used to build models that precisely represent the relationship between effort and one or more project characteristics via the use of algorithmic models. Example: COCOMO.

My Findings:

·         Classification of effort estimation mechanism.

·         use of case-based reasoning and regression trees on different projects

Accuracy of an Effort Model?

Measuring the predictive accuracy of an effort estimation model m or technique t is a four-step process,

Step 1: Split the original data set into two subsets: validation and training.

Step 2: Use the remaining projects (training subset) to build an effort estimation model m. can be used explicit model (e.g., case-based reasoning)

Figure 2: Overall process to measure prediction accuracy

Step 3: Apply model m to each project pn to pq, and obtain estimated effort.

Step 4: Once estimated effort and accuracy statistics for pn to pq have been attained, aggregated accuracy statistics can be computed, which provide an overall assessment

Several datasets are used to simulate a situation where a Web company has a subset of new projects

·         Measuring Effort Prediction Accuracy

o   magnitude of relative error (MRE)

§  MRE =

o   mean magnitude of relative error (MMRE)

§  MMRE =

·         Cross-Validation: The splitting of a data set into training and validation sets is also known as cross validation

My Findings:

·         Prediction accuracy - MRE, MMRE, MdMRE, Pred and absolute residuals.

Sizing Web Applications

One of the survey showed some Web measures taxonomy

Figure 3: Web measures taxonomy

Second survey is about Web quality model (WQM). It is structured based on three orthogonal dimensions.

·         Web features

·         Web life-cycle processes

·         Web quality characteristics

This survey measures according to a second set of criteria –

•         Granularity level: Whether the measure’s scope is a Web page or Web site

•         Theoretical validation: Whether or not a measure has been validated theoretically

•         Empirical validation: Whether or not a measure has been empirically validated

•         Automated support: Whether or not there is a support tool that facilitates the automatic calculation of the measure

Size Measures Taxonomy

Taxonomy uses nine different categories to be applied to each size measure identified in the literature. These nine categories are as follows.

a.       Motivation

b.      Harvesting time

c.       Measure foundation

a.       Problem-orientated measure:

b.      Solution-orientated measure

d.      Class

a.       Length

b.      Functionality

c.       Complexity

e.       Entity

f.       Measurement scale type

a.       Nominal

b.      ordinal

c.       interval

d.      ratio

e.       absolute

g.      Computation

a.       Page count

b.      Connectivity:

c.       Connectivity density­

h.      Validation

i.        Model dependency

My Findings:

1.      Solution orientated and measured length measurement, most of the case.

2.      attributes of Web applications is the main criterion were measured directly using a ratio scale