Tuesday, December 13, 2011

Artificial Neural networks in marketing

What is an Artificial neural network?

Human brain is made of large number of nerve cells or neurons. These are interconnected in a beautiful fashion to form what is called the Neural network. This neural network is responsible for what we are, what we do in our daily life, the reason we learn, remember, process etc. This is what makes human intelligent.

What if we could simulate the human brain’s neural network? Imagine a computer program that somehow is built in the same manner as the brain is. It could do wonders. Alright I have to agree I am being insane here. It’s not practically possible to simply simulate what our brain does using a computer program. But yes let’s say if we can build a program that is very very very inferior to our brain still it can do a lot of wonders. The main reason is because it can learn and perform everything on its own. This particular software code that tries to simulate the neural network of human brain is what is called Artificial neural networks (Neural networks created artificially) in short ANN.

Why ANN?

Let’s look at the traditional approach to programming. Say we have an event occurring, if a condition ‘A’ is satisfied do something if a condition ‘B’ is satisfied do some other thing. If there are multiple events we write multiple codes for that. Now an undeniable fact is that the programmer should know all the cases and the paths that are going to be taken and should provide a case specific action. What if the programmer forgets a case? Or what if the cases are too much to handle? What if the cases are not known or cannot be thought of? The programmer has to again write the code after the issue is reported and build it again. And this process keeps happening like we get version 1 of software, version 2, 3 and so on.

Now what would it be when you have a system that can learn automatically. If there were to be a program that is intelligent to learn, adapt and perform operations. This is possible using ANN. ANN offers an unique approach to look into a problem. ANN does not require explicit coding of a given problem which gives a significant advantage over the traditional approach. ANN is very flexible, it can be trained, and it can Measure and interpret expectations

Alright!! enough with the technical talk. What has this stupid thing got to do with Marketing?

In simple terms, ANN for Marketing can be viewed as software tool that helps the Marketers to take decisions. Before the actual marketing process, there is something called the market research. First, I will deal with specific examples here for easier understanding. Let’s say for market research you have a survey. Say you have 10 people being interviewed. Fine a human can personally sit down, interpret the answers and to an extent give a quantified output. Let’s say we have 10000 people are being interviewed. Now its not possible for human’s to do it. Even if you employ a large number of human’s to do it the disadvantage is that the way everyone percepts the answer will be different. Secondly, it leads to wastage of human resources. The most obvious choice is to look for a software tool. Let’s use the traditional approach to build a software tool. Obviously this software has to identify patterns of words or text that are to relevant and convert them into something that can be measured or in other words quantified. Now what if the pattern is not known to the programmer? An answer can be in a variety of forms. I can give an answer like – “This is bad”. Another person can say the same thing in a different form – “This sucks”. Now both nearly mean the same thing -that the customer is not satisfied about something. Using a case based approach this becomes really tedious. But with ANN it’s much easier.

ANN can be recognize patterns, pick up key information that are not easily identifiable, develop relationships among them. Nothing is explicitly coded. There are also ANN’s that can automatically interact with the environment for learning without any explicit training.

While doing market research, it is very difficult to develop the relationship among the parameters, recognize patterns. ANN does this automatically. One can even find those missing parameters, check their importance and much more.

One more interesting capability of ANN is its predictability. What happens next is a big question mark? Given the intelligence of ANN, it is known to have a very good predictability. Say how a price of commodity would be at a later stage, is a key factor for building the marketing strategy. ANN’s predictability in many cases have proven to be close to accurate.

For more resources, please post a comment with your mail id.

0 comments:

Post a Comment