Artificial Neural Networks ?

What are Artificial Neural Networks (ANN)? : ANN's are programs characterized by a massively parallel but highly interconnected architecture.  The strength of an ANN is pattern recognition and pattern classification, but these programs can also be used for predictive purposes.  They use a series of neurons in what is known as the hidden layer (middle layer) that apply nonlinear activation functions in the data.  ANNs are used in various fields such as industrial process control, speech recognition, financial market forecasts and chemical compounds identification, to name a few.

Two types of neural networks are used in data mining

1.    Supervised: NN is presented with a target variable and fits a function to predict the target variable. Alternatively, it may classify records into levels of the target variable when it is categorical, analogous to statistical procedures such as linear and logistic regression

2.    Unsupervised: NN does not have a target variable, but finds characteristics in the data, which can be used to group similar records together, analogous to cluster analysis

ANNs vs. the common Regression. Some pros......

·         ANNs present the main advantage of not being based on “a priori” assumptions, required in the conventional statistical techniques. Below are a few such assumptions

·         The predictors are linearly independent, i.e. no predictor can be expressed linearly in terms of any other (Multi-collinearity)

·         The variance of the error is constant across observations (Homoscedasticity)

·         The errors terms are uncorrelated

·         ANNs can fit data where the relationship between independent and dependent variables is unknown

·         Self-organized neural networks can learn and update, based on the new data automatically

.....and some cons:

·         ANN Model is generally a ‘Black Box’ i.e. neither the relationship nor the strength of the relationship between dependent and the independent variables is revealed

·         ANNs sometimes tend to over-fit the given sample of training data

The final verdict :  With all its advantages, there are still limits to where the ANN could be put to use.

USE: Where the prediction is the end result and the ANN can adapt itself to the changing customer behavior. Eg Credit Scoring, Fraud Detection, etc.

DON’T USE: Where the output needs to be supported by the relationship of the independent variables to the dependent variable. Eg Marketing Campaign Targeting, Response model, CRM etc. 

Optimizing Acquisition Investments in the Insurance Sector

Insurance industry belongs to the category of businesses where Analytics can help increase efficiency in each phase of the customer life cycle – from acquisitions to cross-sell to retention. Traditionally, the underwriting process was the one with maximum analytics exposure (and the realm of actuaries).  But with high claims ratio and rising acquisition costs, the inclusion of analytics in other areas has become indispensable. Within the whole cycle, the customer acquisition process stands to gain the most with the help of Analytics. 

Acquisition costs range from 5% to 25% of the premium for most Insurance firms. The scope of optimizing these costs using analytics is high and can be implemented in following steps:

1.    Building a robust data management system: Information from internal and external sources should be collected and archived. Large organisations often have structured and unstructured information, spread across various platforms. It is essential to maintain cross-organisational data, along with data from external sources like credit bureaus, Govt. agencies, demographic and psychographic sources, etc. in order to harness the true capabilities of analytics.

2.    Preparing prospective customer’s data: For completely new leads, external data would be the major source of analysis. This would include data from application form, as well as third-party sources like Acxiom, credit bureaus (if legally permissible), etc.

3.    Developing the analytical framework: There are various modelling techniques like Logistic/ Linear regression, Neural Networks, Segmentation using CHAID/ CART etc., available in order to develop the appropriate analytical framework. These frameworks should be able to classify customers according to their probability of conversion.

4.    Optimizing the expenditure: On the basis of probabilities and classifications obtained by using the analytical framework, acquisition investments can be strategically channeled in order to maximize returns.

Marketelligent recently helped a US Insurance provider reduce its acquisition cost by 25%, allowing for a significant impact on its profits. Please refer to this link for details.

Marcia Tal Joins Marketelligent Board of Advisors

Marketelligent to benefit from Marcia’s extensive expertise in Decision Management

New York, NY (Aug 6, 2012) - Marketelligent, a leading data analytics services provider, today announced that Marcia Tal, Founder, Tal Solutions Inc., has joined Marketelligent’s Board of Advisors.  Marcia, a seasoned executive, will bring business leadership experience from the Decision Management space to Marketelligent.

"We welcome Marcia to the board and expect to benefit from her valuable experience leveraging data for better business performance," said Dr. Anunay Gupta, co-founder and Chief Operating Officer of Marketelligent, Inc. "We anticipate that Marcia will provide an important perspective and sound guidance as we continue to grow our business."

"Joining the advisory board at Marketelligent presents an exciting opportunity for me," said Marcia. "Marketelligent is delivering significant value across industries, with its in-depth expertise in data analytics. I look forward to participating on the Board of Advisors, and leveraging my professional experiences to impact Marketelligent’s growth."

Prior to her founding Tal Solutions, Marcia spent 25 years at Citigroup, where she was recognized for creating and building Citigroup’s Decision Management function.  Marcia was the EVP for Decision Management, where she institutionalized an analytics driven mindset across business decisions.