Data mining techniques applied to medical information.

Knowledge discovery from the dramatically increased data of an auto-stored medical information system is still in its infancy. The purpose of this study is to use widely available and easily operated techniques that can satisfy general users in extracting specific knowledge to make the medical information system more functional. Data mining techniques, including data visualisation, correlation analysis, discriminant analysis, and neural networks supervised classification, were applied to heart disease databases. These techniques can help to identify high risk patients, define the most important factors (variables) in heart disease, and build a multivariate relationship model to show the relationship between any two variables in a way that such relationships are easy to view. Simple visualization techniques were utilised to construct this model, which corresponds with current medical knowledge. Two nonparametric (distribution assumption free) classification tools were employed to identify high risk heart disease patients. Both the neural networks supervised classification methods and the discriminant analysis method produced reliable classification rates for heart disease patients. However, neural networks yielded a higher percentage of correct classifications (averaging 89%) than discriminant analysis (79%). Data visualisation and correlation analysis resulted in similar conclusions regarding the most important factors in heart disease. These data mining tools provide simple and effective methods of extracting knowledge from general medical information. The treatment of missing data is also discussed.

Data strip mining for the virtual design of pharmaceuticals with neural networks.

A novel neural network based technique, called "data strip mining" extracts predictive models from data sets which have a large number of potential inputs and comparatively few data points. This methodology uses neural network sensitivity analysis to determine which predictors are most significant in the problem. Neural network sensitivity analysis holds all but one input to a trained neural network constant while varying each input over its entire range to determine its effect on the output. The least sensitive variables are iteratively removed from the input set. For each iteration, model cross-validation uses multiple splits of training and validation data to determine an estimate of the model's ability to predict the output for data points not used during training. Elimination of variables through neural network sensitivity analysis and predicting performance through model cross-validation allows the analyst to reduce the number of inputs and improve the model's predictive ability at the same time. This paper illustrates this technique using a cartoon problem from classical physics. It then demonstrates its effectiveness on a pair of challenging problems from combinatorial chemistry with over 400 potential inputs each. For these data sets, model selection by neural sensitivity analysis outperformed other variable selection methods including forward selection and a genetic algorithm.