A novel approach to uncover the patient blood related diseases using data mining techniques

Abstract

A medical record in general is a systematic documentation of a single patient's long-term individual medical histoiy and treatment. In medical field, patient records are used for analysing their health problem. Clinical dataset is the essential medical record which deals with patient's health details. In the traditional method, medication can be provided to only one patient at a time and it is difficult to identify group of the people having similar symptoms. Multiple health assessment is time consuming and impractical. The present study proposes a new methodology to find potential information related to blood oriented diseases. Generally, real world Complete Blood Count data are susceptible to noise and not suitable for computation. So, there is a need for data pre-processing. Among the refining techniques, data transformation method such as normalisation and data recoding are applied on relevant attributes of Complete Blood Count. Grouping of people having similar health problems can be done by unsupervised learning. Expectation Maximization Clustering algorithm and k-Means clustering algorithm together clusters effectively the patients based on the attributes. It is shown that refined data produces optimum result and may be useful for medical community to diagnose a group of patients.