Pablo Rodríguez-Soria

Analogies and Differences between Machine Learning and Expert Based Software Project Effort Estimation

This paper presents a review and comparison of the software project cost estimation methods that have emerged with more impact in recent years; Expertise and Machine Learning methods. These methods and models have been selected according to an own criteria focusing onto Analogy estimation models and Case Based Reasoning approaches, assuming that they are widely utilized by researchers and with good accurate results. Finally we show a comparative analysis of the seven models proposed inside the Machine Learning methods with advantages and disadvantages between them.

Assessing the Documentation Development Effort in Software Projects

From the initial stages of software engineering, one of the most important practices to be carried out during the software development is a good documentation generation. Since then, this has become more and more important in the overall process of software production of any company, especially for those that have or are trying to achieve higher maturity levels. So for those organizations with a maturity level higher than CMMI level 2, or those that have to comply with the IS0 9000-3 standard, the elaboration and revision of all the components included in the project documentation need an appreciable effort from the development teams. This means that the effort estimation models will adjust this effort driver, to the most accurate precision, in order to obtain correct estimates, which will be used generically and in local environments. In order to do so, we have defined an experiment with the following objectives: To obtain the relationship between documentation effort and total development effort and to obtain updated factors for software documentation, according to the latest documentation standards and software development techniques.

An experimental study on the conversion between IFPUG and UCP functional size measurement units

The use of functional size measurement (FSM) methods in software development organizations is growing during the years. Also, object oriented (OO) techniques have become quite a standard to design the software and, in particular, Use Cases is one of the most used techniques to specify functional requirements. Main FSM methods do not include specific rules to measure the software functionality from its Use Cases analysis. To deal with this issue some other methods like Kramer’s functional measurement method have been developed. Therefore, one of the main issues for those organizations willing to use OO functional measurement method in order to facilitate the use cases count procedure is how to convert their portfolio functional size from the previously adopted FSM method towards the new method. The objective of this research is to find a statistical relationship for converting the software functional size units measured by the International Function Point Users Group (IFPUG) function point analysis (FPA) method into Kramer-Smith’s use cases points (UCP) method and vice versa. Methodologies for a correct data gathering are proposed and results obtained are analyzed to draw the linear and non-linear equations for this correlation. Finally, a conversion factor and corresponding conversion intervals are given to establish the statistical relationship.

Improve Tracking in the Software Development Projects

A proper planning is one of the main issues the success of a software project lies on. However, a perfect planning is not useful without monitoring sufficiently, which involves not only controlling that the activities and expected costs are being met properly, but also having the ability to anticipate the impact that deviations from the plan will have on the future development of the project. Monitoring methods that can be found in literature, can obtain stationary photos of the status of a project at any given time of its evolution. They also allow planning which will be the result in costs terms and final project time obtained from these stages points, though they do not tell us how the project will evolve from the time when we are auditing to completion. By using the family of curves called sigmoidal or S curves, parameterized properly and using monitoring technique EVM (Earned Value Management), the project evolution is mathematically modeled as a continuous function, which allows not only getting a picture of it but also a movie that show the full evolution to its final draft. In order to validate the proposed function, show will use a set of projects in the area of software development with different characteristics that allow the proposal address to face up to a wide variety of situations. All these projects come from the real worldwide, and we proceed with two experiments these projects come to validate the proposals of the paper.

Early Functional Size Estimation with IFPUG Unit Modified

Nowadays functional size measurement is a strategic key to deal with the management of software systems development. The origin of this importance is the fact that functional size measurement is the main input variable in software effort estimation systems. Nevertheless, to obtain precise functional size measurements it is not only necessary to have a lot of information of the system to be developed, but also software project planning is one of the early stages in the project. To solve this difficulty, one of the main software management research technique is centered in the study of methods to obtain precise functional size measurements early in the development phase for early functional size estimation. The functional size unit selected to do the study has been IFPUG because is the most widely used method.

COSMIC Measurements Dispersion

Over the years, software functional size measurement has become a viable tool to determine the effort and time needed to carry out a software project. This growth has led to an increased interest, as shown in several studies, as well as to the development and optimization of functional size measurement units. IFPUG Function Points since their development have been the most widely used functional size estimation approach. However, during recent years, a new method, called COSMIC FFP, has emerged as a second generation unit for functional software measurement, and has brought new features and benefits. The aim of this paper is to study the error that is introduced together with the interpretation of the unit application rules, thereby focusing on COSMIC CFPs. Due to subjectivity and degrees of freedom, this error can lead to great measurement dispersions. We have analyzed these dispersions, based upon a study that we have performed at our university, together with a group of Master students.

Visual learning techniques for software measurement

In Computer Science, particularly in Software Engineering and Software Measurement, the level of abstraction and complexity of some concepts is very high. That implies a difficulty for the students when they try to assimilate those concepts because in many cases they cannot make mental associations between the idea and a real image. With the development and use of new learning techniques such as visual learning, the assimilation of these software measurement concepts could be facilitate by the use of a more intuitive way than traditional resources, improving learning results and in consequence academic results. To validate this hypothesis, CuBIT Software Measurement Laboratory developed a software functional size visual course and carried out an experiment at the University of Alcalá, Spain, comparing the traditional teaching methods with the use of these visual learning techniques. The results of this experiment, presented in this paper, indicate that the use of visual learning techniques could improve the learning process.

Horizontal dispersion of software functional size with IFPUG and COSMIC units

Software development companies today are widely using software functional size measurement (FSM) as the main variable to assess the effort and time needed to perform a new software project. In the recent years, this has led to a grown interest in improving the way the measures are taken. In such sense, one of the main aspects that could have impact on measurements and that has not been enough studied is the error introduced by the measurer of the software application, through the subjectivity that can be introduced in the interpretation of the unit application rules. Such error could be evident in a measurement dispersion, defined in this paper in two possible ways: (a) Horizontal Dispersion, where the error could be introduced by the fact that two or more different people counted the same application at the same moment in the project development; and (b) Vertical Dispersion, where the error could be introduced by same measurer that count the same application at different times during the development. Since its definition by Albrecht in 1979 and its subsequent change of name in 1986, IFPUG function points have been the functional software measurement unit mostly applied, despite the definition and standardization of other variants such as NESMA, Mk-II, or more recently FiSMA. However in recent years a new method has been introduced called COSMIC that has been defined as a 2nd-generation FSM method, attracting the interest of the international software measurement community. The aim of this research is to draw some preliminary conclusions from statistical analysis of the software functional size data in which the horizontal dispersion degree could have been introduced in measurements taken into account IFPUG and COSMIC methods.