TECHNOLOGY OVERVIEW The core intellectual property holding of Machine Insight is currently ECS, a new machine learning algorithm that dramatically outperforms other machine learning algorithms on certain classes of difficult problems. ECS has the following performance advantages: ECS is a rule induction system, and generates rules that can be understood by people. This is important for human understanding of the patterns found by the system, and will allow domain experts to validate the wisdom of acting on these patterns. ECS search mechanisms are problem and data format independent. ECS can be broadly applied to many different kinds of problem. ECS minimizes the number of rules required to describe a condition, without loss of accuracy. This is critical when the rules the system generates are interpreted by people in order to gain insight into their data. Finding the fewest possible rules without loss of accuracy is known as generalization, and ECS has excellent generalization properties. Optimization is very fast. ECS finds solutions to benchmark problems in seconds that other machine learning techniques cannot solve in any reasonable time frame. ECS can handle problems that are many orders of magnitude more complex than other techniques. ECS works effectively with very noisy data. ECS measures the certainty of its predictions as a core part of the search and optimization processes. This helps prevent the classic machine learning problem known as over training, and also assists in providing the outstanding generalization properties of ECS. ECS is effective at rejecting features in large noisy data sets that are random artifacts instead of real signals, again as part of the core optimization process, even though these artifacts may appear to be significant by standard statistical measures. This helps alleviate the machine learning problem known as data snooping. ECS is effective on non-stationary problems, where aspects of the problem are changing even as predictions are being made. This can be useful when data arrives incrementally, such as with control problems or intra-day financial data.