A recent paper by Keith R. Bisset and his team at Virginia Tech offers an exciting new software engine for modeling epidemics. Named “Indemics” (Interactive Epidemic Simulation), the software enables public health policy analysts and epidemiologists to analyze the spread of diseases throughout the population. While a noteworthy accomplishment in and of itself, the software could also be used for interactive simulation of a marketing campaigns for a national brand.
Indemics builds on a wide foundation of previous work in epidemic simulation. Traditional systems were based on a series of ordinary differential equations, however these have transitioned to individual-based simulations over the years. These simulations model individual people, and how they are infected as they make contact with other individuals in their daily lives. The simulator models the spatial location and interactions between people, such as at work, school, the gym, etc. Then, along with each interaction, there is a probability that an infected person will pass the disease on to someone else.
The “world model” in modern epidemic simulations is intensely detailed, with socio-economic data influencing the probability of the spread of disease, and geo-spatial data showing how sparsely populated rural regions spread disease differently from urban centers. The latest census data is used to keep the models up-to-date, along with LandScan data for the geographic information.
What Indemics brings to the table is more flexibility in modeling the response to disease, as well as a relational database back-end. While previous work assumes that the response to the disease will be uniform, or adhere to a set of pre-defined rules, Indemics allows for a fully-decoupled representation of the disease treatment and public policy influence. The relational database back-end, on the other hand, allows for ease of development, both in the modeling and reporting of results.
These new developments allow the project to be easily extrapolated to other domains, such as marketing. While Big Data offers great opportunities for analyzing the performance of past campaigns, predicting future campaign performance is made difficult through the complex, dynamic nature of the system. By instead modeling the full system in an individual-based simulator, the software would be able to accurately predict not just the effectiveness of direct marketing sources, but the effect of word-of-mouth and indirect marketing as well. Connected with the census data, the simulation could be used to find under-served markets ripe for expansion, as well as provide alerts of possible areas of over-saturation.
The primary challenge in the simulation would lie in computing the probability of product purchase based on the demographic and geo-spatial data. For this, traditional techniques would need to be used to analyze the historical information and back-propagate weights based on marketing campaign vs sales data.
While applying epidemic simulations to big data marketing analysis offers significant promise, it does require a large-scale implementation for accurate modeling. Implemented properly, however, the simulation could provide truly fascinating insights through its predictive analysis of marketing campaign effectiveness.
Written by Andrew Palczewski
About the Author
Andrew Palczewski is CEO of apHarmony, a Chicago software development company. He holds a Master's degree in Computer Engineering from the University of Illinois at Urbana-Champaign and has over ten years' experience in managing development of software projects.
Google+
Great article – Did you read the last article from CIO magazine – CIO DATA structure will overcome cancer?