Neural Network: Is it needed for your business?
Dr. Purnima Sangle & Swen Mehta

In this era of globalisation businesses are becoming more and more information based, businesses are becoming more and more information based, businesses are no longer dominated by labour& raw material costs, but driven by information. The core competence of any organization is its knowledge about its business either explicit or implicit-which lies in the database of the organisations. Neural network based products helps businesses to derive this implicit useful information from database with the help two complementary processes, first segmenting the data into meaningful clusters and then predictive modeling.

Neural network the term traces its origin from our biological nervous system, which is made up small identical cells called neurons connected to each other through links called “synaptic connections” and as we experience things, the learning’s involved in these events changes the composition of connections between neural cells, not in the cells themselves. Similarly a artificial neural network is a system loosely modeled on human brain and mimics certain processing capabilities of brain. It is an attempt to simulate within specialized hardware and /or sophisticated software, the multiple layers of simple processing elements neurons. Each  neuron is linked to certain of its neighbors with varying coefficients of connectivity that represents strength of these connections. Learning is accomplished by adjusting these strengths to cause the overall network to output appropriate results. The analogy with brain is shown in diagram :

When approached with a proposal to apply neural network, how should a business manager evaluate the proposal? Does this new capability offer real benefits, or is this the latest example trendy approaches and buzzwords? Most importantly are these techniques practical or are they academic approaches that are not practical and cost effective?

Given the steady increase in successful applications neural networks are for real and offer substantial benefits. Neural networks should be applied in situations where traditional techniques like logical analytical analysis; DSS, expert systems etc have failed to give satisfactory results or where a small improvement in modeling performance can make significant difference in operational efficiency or in bottom line.  Direct marketing is excellent example of where a neural network based software can analyze large quantities of data to establish a pattern and characteristics in situation where logic or rules are not known. A manger interested in getting more useful information from available data should consider neural network technology as an option. These can be used aggressively by organization to focus available resources more effectively thus gaining valuable competitive edge. The applicability of neural network techniques v/s traditional techniques is as shown  in table :

The neural network algorithms have been very successfully used in data mining applications as well. Specific areas of business which have benefited from standard neural network applications are fraud risk prediction (Master card and Visa card both), customer profiling,  airline management, forex rate evaluations, mobile fraud detection etc.

Of the above said areas fraud detection systems have caught fancy of businesses and these have become largest users of neural network technology. A estimate of credit card industry can be taken from facts that total 12 billion transactions are made annually and approximately 10 million or 1 out of every 1200 transactions turns out to be fraudulent. Also 4 out of every 10000 of all monthly active accounts are fraudulent. In the financial industry annual fraud losses are calculated using a measure called fraud basis points. Currently US fraud losses amount to 10 basis points, which means that out of every 1 dollar ,one tenth of cent is lost in fraud. In fraud detection two software packages have become industry standard-FALCON & PRISM. They are based on neural network technology and are designed to interface with current business software for credit card management. The decision engine that FALCON uses is as shown :

In the end looking at the technological advancements and benefits provided by neural network technology it is high time for businesses to adopt these application to make them competitive and profitable.

Article Contributed by:

• Dr. Purnima Sangle (Astt. Professor NITIE, Mumbai)
• Swen Mehta (Student PGDIM VIII batch, NITIE Mumbai )

Dr. Purnima Sangle, Astt. Prof Information Systems
Swen Mehta, Student of PGDIM VIII, swen_mehta@im8.nitie.edu
NITIE, Vihar Lake, Mumbai, 400 087