Easy-to-Use Classifier for Complex Problems

The BDRA is a portfolio of algorithms developed for automated, real-time prediction. The technology not only reduces large amounts of data and information to an easily handled set to support decision-making, but also points out the critical information that should be used. The technology’s use of feature selection and AI techniques allows for continuously improved classification performance. Designed for complex problems, BDRA is particularly robust in cases where the patterns in the data are not readily evident or when the number of features in the data set needs to be reduced. BDRA approaches classification and dimension reduction in a fundamentally different theoretical manner than other classification and neural network techniques. Key Advantages compared to other classification methods such as statistical methods and neural networks include: Ease of use. BDRA can be run and interpreted by individuals with very little statistical background. High performance. The technology outperforms traditional classification and artificial neural net methods in certain types of complex problems with high dimensionality, or in situations where there is no known built-in structure or functional form to data. Fast. BDRA guides users to optimal decisions by helping determine the most relevant factors. Reduces risk in high risk decisions. The BDRA employs fundamentally different mathematical techniques than either neural networks or linear classification techniques. Thus while the BDRA can compete with the best methods of classification, perhaps one of its main advantages is its ability to be used in conjunction with other methods for decision support and cross validation. This is particularly important in high risk decisions. Applications BDRA was originally developed at NUWC for classification of sonar data; it has outstanding utility when there is a need to identify quickly inherent patterns embedded in large amounts of data such as insurance scoring, fraud detection, customer profiling for direct marketing, medical diagnosis, and financial market modeling.