Quantifying Relevance of Input Features

Identifying and quantifying relevance of input features are particularly useful in data mining when dealing with ill-understood real-world data defined problems. The conventional methods, such as statistics and correlation analysis, appear to be less effective because the data of such type of problems usually contains high-level noise and the actual distributions of attributes are unknown. This papers presents a neural-network based method to identify relevant input features and quantify their general and specified relevance. An application to a real-world problem, i.e. osteoporosis prediction, demonstrates that the method is able to quantify the impacts of risk factors, and then select the most salient ones to train neural networks for improving prediction accuracy.