Automated Project Cost Estimation : Using Analogies

The ANGEL Project

 Developed by Chris Schofield

 

A machine learning approach, estimation by analogy, is the focus of a research project being undertaken at Bournemouth University. Estimation by analogy involves searching for similarities between a target entity, such as a proposed software project, and a set of historical entities of the same class. Each entity is characterised by a number of attributes that are available at the point when estimates are required. These attributes are used to minimise the Euclidean distance in n dimensional space between a target entity and source entities and thus find the closest 'analogies'. Once a historical analogy has been found, the 'actual' values for the attribute (for example cost) that is to be estimated can form the basis of an estimate for that attribute in the target entity.


Online Papers

The following papers present a more in-depth discussion of the analogy approach in relation to cost estimation

Effort Estimation Using Analogy

Presented at the 18th International Conference on Software Engineering, Berlin March 1996

Software Support for Cost Estimation by Analogy

Presented at the European Software Cost Modelling Conference, Rolduc, NL, May1995

Effort Estimation by Analogy: A Case Study

Presented at the European Control and Metrics Conference, Wilmslow, UK, May 1996


ANGELIC Software

Download ANGEL for FREE!!!

A windows beta version of the analogy estimation tool ANGEL-lite is available for installation.

Coming Soon! - A set of instructions.......

Templates and Datamodels

Each zipped up file contains one template (*.tmp) and one datamodel (*.mdb) for use with ANGEL.

Kemerer.zip - Data ( 15 Projects ) collected from a national computer consulting and services firm specialising in design and development of data processing software.

Taken from : Kemerer, C.F. 'An Empirical Validation of Software Cost Estimation Models', CACM, 30(5), pp416-429,1987.

Kemerer Data items :

Actual Effort Actual Project Effort (Measured in Man Months.)
Duration Project Duration (Measured in Months)
 
KSLOC Thousand Source lines of Code
 
UA_FP Unadjusted Function Point Count
 
FP Adjusted Function Points

Albrecht.zip - Data on 24 applications developed by the IBM DP services organisation.

Taken from : Albrecht, A.J. and J.R. Gaffney, 'Software Function Source Lines of Code, and Development Effort Prediction: A Software Science Validation', IEEE trans. on Softw. Eng., 9(6), pp639-648,1983 Albrecht Data items :

Actual Effort Actual Project Effort (Measured in Thousand Work Hours.)
SLOC Thousand Source Lines of Code
 
IN No. of Inputs
 
OUT No. of Outputs
 
FILE No. of Master Files
INQ No. of Inquiries
Function Points Function Point Count

This page was updated by Dan. Snell - 25th January 1997

Please address any enquires to cschofie@bournemouth.ac.uk