Patterns: detect them and act


Posted on: July 11, 2011 by Almudena Alonso Fernandez

Based on years of observation, our Civilization has discovered patterns that have helped our ancestors to make better decisions for centuries: countrymen acted based on the Nature signals (we still use sayings based on this knowledge: Red sky at night shepherd's delight; red sky in the morning, shepherd's warning / March winds and April showers bring forth May flowers), sailors early predicted storms based on a combination of nature phenomena: changes in ocean currents, slight variations in wind intensity and direction, ancient Egyptians based their agriculture on the pattern followed by the river Nile flooding. Of course, accuracy of these predictions was not always as good as desired.

When technology appeared in our lives, it began to provide information to improve our decisions, first in business environments, where technology was primary adopted. But we also base our private decisions on the analysis of available information.

Analyzing historical information (using complex mathematical algorithms to obtain accurate results), we can find behavioral patterns. But, what can we do with these patterns? A mistake that some organizations did in the past was to invest in using predictive analysis to improve their decisions, which results only reached a manager desk. We have to find out how these patterns affect our business, or how important the effect of a certain pattern is.

And then act. Agility is the key in a Pattern Based Strategy. Detecting in time a natural disaster is not enough; the effects can be controlled only if the appropriate actions are quickly adopted. Detecting a fraudulent use of a credit card when it is occurring is not enough; implementing mechanisms to cancel immediately the card is the only way to avoid a more important fraud. As we all are suffering, detecting the coming of a financial crisis (had it been possible, if not a black swan – see Steve Nimmons article “” in Atos Blog ) is not enough, actions to avoid the effects should have been taken in an early stage.

In the last years traditional patterns are becoming obsolete. Are we using the right information to model these patterns? Only data that can be accessed by the standard methods are normally used to obtain the information to seek for the patterns. We should look for signals to detect them somewhere else.

Social media, smart mobile devices and collaboration tools are new ways of communication that have change people behavior. The signals that determine patterns are no more in the traditional information systems. Relationships in social networks, influence blogs, contextual information from mobile devices, on-line purchases, connected cars, smart metering and domotics are some of the sources of information where we now should dig in.

Our challenge is to find these new signals, model the behavioral patterns with them, but also detect them in the right moment so we can act in an agile manner.

Share this blog article


About Almudena Alonso Fernandez

Business Intelligence Manager and member of the Scientific Community
Universidad Complutense de Madrid. She is member of Data Mining International Scientific Committee of Universidad Carlos III de Madrid since 2004, contributor of Cotec Foundation Data Mining Document (for technological innovation) 2005, and, since 2009, member of Atos Scientific Community. She began her professional career in Marketing Research sector, working for companies like A.C. Nielsen, where she was in charge of quality control of Nielsen panels. She joined Atos (then Sema Group) in 1994 were she has carried out several roles in data analysis projects for Atos main clients: Data analysis and data mining consulting projects based in statistical and artificial intelligence techniques (Multivariate Analysis, Cluster k-means, Time series forecasting, simulations, neural networks, association, etc.); Credit Scoring model design for risk evaluation; Data analysis consultancy, modeling, design, development and support (Financial and Marketing Data Warehouses and Data Marts); Business Intelligence tools evaluation, etc.

Follow or contact Almudena