Semantic extraction using neural network modelling and sensitivity analysis

Abstract

We present three methods of using neural network modelling and sensitivity analysis to extract semantics given historical data of a given system. First, neural network modelling and sensitivity analysis is used to determine the decision boundaries of the system. Several test cases including a simple credit rating system are described to illustrate the use and effectiveness of this method. Secondly, it is used for causal inferencing of the system. That is, determining which inputs has the largest effect on the output of the system. In addition, causal inference under different input conditions was tested. This is done by applying the index on a subset of the input data. Typical problems like the decoder and parity are tested and discussed. Lastly it is used to analyze historical data for detection of exceptions or special input cases. In all three methods, the system being studied which is available only in the form of input-output pair data, is first modelled using a neural network. Sensitivity analysis is then applied to the trained network to extract semantics learned by the network.