Track:
Contents
Downloads:
Abstract:
A scalable/parallelizable/distributed method is shown that was originally developed for data mining but which is extended to perform web mining, text mining, mining of time-series or temporal data, mining mixed-type data, and mine high-dimensional data. It is implicitly stochastic thus it can incorporate statistical methods such as Bayesian. It is based on the automatic generation of assocation rules. Since it also uses a multiplicative fuzzy logic the neural network tuned to the data is easily comprehensible. It lends itself to visualization, and interactive exploration. Since it uses a unique hashing algorithm (for storage and data manipulation), one that works with associative access, it can be used for other techniques such as nearest-neighbor methods. It is shown here that it is a perfect solution for chance-event detection and processing. It lends itself to hardware acceleration and thus can be used for very large scale projects such as streaming data for "homeland defense." The method is an ideal platform for integration of data warehousing, OLAP and data mining.