| Technology Overview Genetic Algorithms (GA) are a specific type of Artificial Intelligence (AI) which is particularly well suited to analyzing large and extremely complex databases. While the many industries have employed advanced analytical techniques (e.g. covariance, regression analysis, K-means testing) for decades and more recently less sophisticated AI (e.g. Neural Networks), there are still significant shortfalls in the research information available to IT consumers and a vast expanse of useful data never employed to increase investment profitability and success. The reasons for this shortfall are relatively straight-forward. In the case of conventional analytics, the sheer volume of data overwhelms available analytical resources—computational cost grows exponentially with every factor included for analysis. State of the art analytical techniques are stretched to consider 10 factors simultaneously--so there just isn’t enough computer hardware to answer every question. Neural Networks offer the ability to examine a larger number of variables simultaneously, but Neural Networks are suited to detecting a single relationship (model) within a data set. Unfortunately, many businesses tend not to be driven by a single overarching relationship but rather the synthesis of 100’s of competing trends and relationships, and hence the relatively lack-luster performance of Business Intelligence products in many settings. It is important to understand that there are many forms of Artificial Intelligence, much the same as there are many types of human intelligence—each suited differently to different tasks. A Genetic Algorithm, unlike other AI techniques, attempts to detect information using a family of models, which continually compete and evolve as more information is gathered. Relationships are automatically prioritized and examined based on the probability that they will yield information value. This means that not only are “questions” asked and answered at machine to machine speed, but a GA only asks those questions it believes will hold and information “payoff” – drastically reducing the computational expense of your research (and allowing the GA to tackle significantly more challenging problems with less computer hardware). GA’s are therefore more resilient when attacking tough, complex models, and significantly less affected by noisy, “dirty” data. The strength of the GA technology, combined with DaMi2’s unique, Patent Pending GA architecture, allows us to routinely analyze 50-100 variables simultaneously on a relatively small hardware platform. While the technology is fairly complex, the principle is relatively straight-forward—the more factors you can practically examine, the more precise and productive your projects can be--and DaMi2 can look at 20-30 times more stuff than other competing technology. |
