Pattern Recognition in Medical Decision Support and Estimating Redundancy in Clinical Text

Abstract

Feature sets selected for multidimensional pattern classification are estimated using a novel criterion. The priority of features for each class is generally optimized to maximize relevancy and minimize redundancy between each class. While mutual information can be used to estimate relevancy information, redundancy information cannot be estimated since its dynamic ambit is determined by feature and class. In addition to assisting in adjust of changing patterns, support vector machines are helpful in classifying normal and abnormal patterns. In pattern matching, Random Forest Logic's pattern recognition algorithm lends itself easily to pattern matching algorithms. An algorithm for evaluating the classification performance of health care medical data is proposed. A variety of experimental results confirm that the proffered technique is more veracious than conventional algorithms relating to the classification of accuracy when the number of selected features is varied.