Artificial Neural Networks for Business Analytics

Artificial Neural Networks for Business Analytics

Copyright: © 2014 |Pages: 16
DOI: 10.4018/978-1-4666-5202-6.ch019
(Individual Chapters)
No Current Special Offers

Chapter Preview



Humans are naturally suited for recognizing and interpreting patterns; however, large and complex datasets, as in Big Data, preclude efficient human analysis. Computational pattern recognition encompasses means of describing, classifying, grouping, categorization, and/or clustering data (Jain, Duin, & Mao, 2000). Expanding on the concept of pattern recognition is knowledge discovery in datasets (KDD), the process of finding non-obvious patterns or trends in datasets primarily to assist in understanding complicated systems (Mannila, 1996). Within a world of decreasing cost of computer storage, increasing capabilities of computer processing performances, and increasing complexity of today’s business problems to solve, data analysts are encouraged to adopt more robust data mining methods. One such methodology described in this chapter is an artificial neural network (ANN).

Perspective of Chapter

ANNs provide one approach of solving complex problems in business (Jain, Duin, & Mao, 2000). ANNs are desirable because they provide a well-structured framework to discover non-linear relationships within data sets that are considered ‘noisy’ or complex. Thus, the primary benefit of utilizing this type of model is the improved accuracies that are obtained when creating models for classification or prediction, which can be used to make better business decisions. For example, business decisions related to up-selling, cross-selling, or demand-planning could all be improved upon if more accurate decision models are created. Though improved accuracies can be obtain, there are many reasons why the field of business has been slow to adopt the modeling practice. For example, ANNs are rooted in machine learning, which may be intimidating for some, especially, since the term is relatively new. ANNs also require much more data to derive than common methods like multiple linear regression. They also take a great deal of time to develop, since a trial-and-error approach is needed to determine the ‘best’ model. In addition, while ANNs have proven to be accurate classifiers and predictors in business applications, some hesitation and misconceptions appear due to the ‘black-box’ nature of ANNs (Dewdney, 1997) (de Marchi, Gelpi, & Grynaviski, 2004), hence traditional statistical-based models are far more used in practice.

Objectives of Chapter

The objective of this chapter is to provide readers with a general background of ANNs and their business applications, where the target audience is intended to be readers who may not be familiar with this form of mathematical modeling practice but may want to pursue it for their business needs. The goal of this chapter is to provide readers with a background and common issues related to the development of an ANN. As a starting point, this chapter begins by describing the foundational concepts that relate to ANNs. Moreover, an overview of their biological inspiration, their mathematical representation, and architecture will be discussed in an effort to make the topic less intimidating to readers. Additionally, the authors also provide a brief overview of the primary steps that are needed to derive an ANN model. For instance, data needs to be pre-processed, an ANN architecture needs selected, training methods and heuristics need to be implemented, and the results need analyzed. Finally, the authors have provided readers with a list of business related endeavors as a suggestion for further reading.


ANNs refer to interconnected networks of nodes, which manipulate data from input to output feature-space (Jain, Duin, & Mao, 2000). ANNs are computationally complex and the learning curve to develop an ANN use to be very step; however, various software packages are now available for practitioners, which remove much of the burden placed on truly understanding the mathematical backbone in which ANNs are derived. For example, titles include SPSS (2013), which is perhaps the most well adopted data analysis software package, NeuroDimensions (2013), which has an extremely powerful, yet user-friendly add-in to Microsoft Excel, and other options such as Matlab (2013), and R (2013).

Key Terms in this Chapter

Over-Fitting: Occurs when a mathematical model describes random error or noise instead of the real underlying relationships within a dataset, which artificially produces desirable goodness of fit metrics for training data, but produces poor metrics for testing data.

Knowledge Extraction: The process of discovering how input attributes are used within an ANN to formulate the output such that one can validate functional relationships within the model.

Unsupervised Learning: A learning strategy of developing an ANN in which the desired output, or dependent attribute, is unknown.

Post-Processing: A process of utilizing a trained mathematical model in order to improve the understanding of the database that has been modeled.

Epoch: The representation of an entire training set of sample data through the learning algorithm so that an ANN’s weights can be determined.

Back-Propagation: A supervised learning method used to determine the weights of an ANN, where the difference between the desired and the model’s output is minimized.

Pre-Processing: A process of preparing a dataset in order to develop a mathematical model.

Supervised Learning: A learning strategy of developing an ANN in which the desired output, or dependent attribute, is known.

Neuron: An individual building block of an ANN in which weighted input values are transformed via a transfer function into an output, which is typically passed to other portions of the network.

Complete Chapter List

Search this Book: