A Distributed Approach to Finding Complex Dependencies in Data
Matthew D. Schmill · 1998
Learning complex dependencies from time series data is an important task; dependencies can be used to make predictions and characterize a source of data. We have developed Multi-Stream Dependency Detection (msdd), a machine learning algorithm that detects complex dependencies in categorical time-series data. dmsdd strives to balance the search for strong dependencies across a heterogeneous network of workstations. We develop a load balancing policy for dmsdd-- first using only static techniques, and then adding in dynamic measures -- on canonical machine learning datasets. Keywords: prediction, dependency detection, time series, distributed algorithms, load balancing 1 Introduction We are concerned with the analysis of dependencies in time series data. Examples of this kind of data include economic indicators, distributed network status reports, and binned continuous streams such as flight recorder data. A successful data mining technique might elicit many useful details from such d...