Golay Code classifier approach for medical diagnosis

As Precession Medicine applications have led to various forms of new treatment and discoveries, medical data classification problem became a notable data mining challenge. Clinical diagnostic employing machine learning systems is a complicated process as it requires high accuracy and speed. Although, several researchers have tackled this problem, big data poses some difficulties on achieving the desire accuracy and speed. This challenge increases substantially as the multidimensional datasets loaded several hurdles on the current computational model. We suggest using a hash coding method to tackle this problem. In this work, an innovated supervised method that reverses the error-correction Golay Code to classify large scale datasets and predict individual patient outcomes or risks based on a chosen number of features (i.e. medical finding) is presented. While some classifiers are very flexible, with many user-adjustable parameters, the proposed algorithm is almost automatic. Its cost to classify unseen pattern is on average O(1). To support our theoretical arguments, empirical evidence is considered. Different classification algorithms are also discussed and reviewed for their similarity. Our experiment results show that the proposed approach is able to achieve a good generalization performance, compared to the results of other classifiers.