A VectorBased Approach to Software Size Measurement and Effort Estimation

1
A Vector-Based Approach to Software
SizeMeasurement and Effort Estimation

• Area Software Metrics
• Authors T.E. Hastings and A.S.M. Sajeev
• Presented by Rajya Lakshmi Kanchumarthy
• IEEE TRANSACTIONS ON SOFTWARE ENGINEERING, VOL.
  27, NO. 4, APRIL 2001

2
INTRODUCTION

• Software size
• Addressing both functionality and problem
  complexity.
• Vector Size Measurement
• Vector Prediction Model

3
Purpose of VSM VPM

• VSM
• To measure the size of software systems.
• To classify software systems.
• VPM
• To estimate development effort early in the
  software life cycle.
• To manage software engineering economics by
  careful pre-planning.

4
Software size Measurement

• We measure processes, products and resources.
• -process-s/w related activities
• -deliverables such as requirements,
  design,
• code etc.
• -assets like people, equipment, tools
  etc.
• Measured in terms of attributes, entities
  (external and internal).
• Length, functionality and complexity are
  parameters of software size function.
• -length is the total number of entities
  in the system.
• Sfs( l, f, c)

5
Software specification

• To give a concise specification of what a system
  provides.
• -we can make abstractions like use
  cases, data
• modeling, algebraic specification,
  formal
• specification etc.
• -Host other techniques including
  natural language.
• It is an early product of SLC to predict the size
  of the system.
• Ensure that all fundamental entities, their
  attributes and the relationships are specified in
  the specification.

6
Algebraic Specification

• It is a combined and unified approach that gives
  formal mathematical description of software
  taking data and functional aspects of the
  software system and without over specification.
• Algebraic specification is based on Abstract Data
  types.
• Meyer-A class of data structures not by an
  implementation but by the list of services
  available on the data structures and the formal
  properties of these services.
• Composed of Functions and Rules.
• - ADT F . R (where, F-gtSyntactic
  R-gtsemantic)
• -Rules are set of assertions and axiom
  that specify
• semantics.
• -Assertions are specified by a set of
  predicates, pre-conditions
• and post-conditions.
• The specification of a system (subsystem or
  component) is the set of all ADTs.
• SystemADT1,ADT2,..ADTn

7
ASL

• Algebraic specification is based on ASL.
• ASL ( Has no reserve words but has few
  symbols).
• An ASL specification is composed of a
• set of abstract data types (ADT).
• Predicates, axioms and assertions.
• -A predicate, is a mathematical relation in
• Maths.

8
Vector Size Measurement

• Software size measure in terms of vectors has two
  attributes. They are Functionality and problem
  complexity.
• VSM is represented by ADTs.
• Representation-operators, operands.
• Calculating f, c, l (eg- bank account).
• Size, SF (f , c)

9
Vector representation

• Vector representation in 2D.
• Has both magnitude and gradient
• Calculating m and g.
• m takes complexity and functionality into
  account in a balanced and orthogonal manner.
• g indicates the relative dimensions
  (characteristics) of the software system.

10
Vector prediction Model

• The purpose of VPM is to predict the effort
  required to produce a solution.
• Methods-gt expert opinion, analogy, decomposition,
• and models.
• VPM is model based.
• Productivity Based Prediction Models(PS/E)
• Linear Regression (Eab S,is based on least
  squares)
• COST models (Ea pow(S,b)(F)). Eg COCOMO.
• -F Effort adjustment factor
• -Calculated as product of all cost
  drivers in a model.
• -primary factors-size of system measured
  in functional points
• -secondary-application type, team size,
  team experience, project duration etc.
  
• Vector approach

11
Formal Validation

• For an attribute to be measurable, it must allow
  different entities to be distinguished from one
  another.
• A valid measure must obey the Representation
  Condition.
• - Representation condition To measure
  the attribute we need to have corresponding
  relations in some number system then measurement
  is the assignment of numbers to the entities in
  such a way that these relations are preserved.
• Each unit of an attribute contributing to a valid
  measure is equivalent.
• Different entities can have the same attribute
  value (within limits of measurement error).

12
Empirical Validation

• Scope
• Objectives
• Methods
• Test Criteria
• Constraints

13
General Discussion

• Uses of VSM/VPM
• VPM approximation
• Benefits of vector approach
• Limitations
• -Extra effort for effort estimation.
• -The size of the project may grow or
• shrink after the estimation is done.
• -A specific system is modelled
  differently by different
• people.So these effort estimation
  techniques give
• varied results.No consistency can be
  achieved.
• -Certain classes of problems may not be
  specified
• adequately using ASL.

14
Summary

• The proposed vector representation provides a
  balanced view where software size is measured in
  terms of magnitude and gradient.
• Theoretical validation confirms that the approach
  does not violate any necessary properties of
  measurement theory.
• Empirical validation confirms that the approach
  can be used early in the SLC for the prediction
  of effort required to provide solutions across a
  range of application types and application
  domains.

15
References

• www.ieee.org
• http//citeseer.nj.nec.com/364985.html -gt
  Software engineering economics.
• http//portal.acm.org/citation.cfm?id381132dlAC
  Mcollportal -gt
• Same Paper published in ACM.
• http//www.cs.usask.ca/grads/vsk719/academic/856/p
  roject/node24.html -gt
• Code Size Measurement.

16
Thank You