Introducing Artificial Neural Networks through a Spreadsheet Model

ABSTRACT Business students taking data mining classes are often introduced to artificial neural networks (ANN) through point and click navigation exercises in application software. Even if correct outcomes are obtained, students frequently do not obtain a thorough understanding of ANN processes. Thi...

Full description

Saved in:
Bibliographic Details
Published in:Decision sciences journal of innovative education 2012-10, Vol.10 (4), p.515-520
Main Authors: Rienzo, Thomas F., Athappilly, Kuriakose K.
Format: Article
Language:English
Subjects:
Analytics
and Data mining
Artificial Intelligence
Artificial neural network
Business education
Business intelligence
College Students
Computer Networks
Courseware
Educational Practices
Educational Strategies
Information Processing
Navigation (Information Systems)
Neural networks
Software
Spreadsheets
Studies
Teaching Methods
Teaching Models
Technological Literacy
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:ABSTRACT Business students taking data mining classes are often introduced to artificial neural networks (ANN) through point and click navigation exercises in application software. Even if correct outcomes are obtained, students frequently do not obtain a thorough understanding of ANN processes. This spreadsheet model was created to illuminate the roles of the following ANN parameters: weights, learning rates, threshold functions, and transformation functions. The spreadsheet ANN model project is given early in the semester, just after ANN is introduced. Students can see effects of ANN parameters as they make changes to spreadsheet model inputs, greatly enhancing discussion of ANN processes. After working with the spreadsheet model, students have expressed an appreciation for decisions based on patterns of historic data, and they like the ability to peek “behind the curtain” at processes of predictive software packages.
ISSN:1540-4595
1540-4609
DOI:10.1111/j.1540-4609.2012.00363.x