Novel Semantic Discretization Technique for Type-2 Diabetes Classification Model

Semantic discretization, which is relatively a new concept, can be viewed as the discretization technique that uses the semantics of the data along with its value. The semantics of the data refer to the domain knowledge inherent in the data. The semantics of data is derived from the data value itself. Objective and context of the study also contribute significantly to identifying semantic of the data. Since no explicit ontology is associated with the data in semantic discretization, identifying, interpreting, and exploiting, the semantics of the data is a challenging task. This paper presents a novel algorithm for semantic discretization, in which machine learning techniques such as classification and association rule mining is used to derive semantic knowledge, which is further used for discretization. To show the effectiveness of the proposed semantic discretization algorithm, we applied it on diabetes dataset. Experimental results show 2–15% improvement in classification accuracy on semantically discretized dataset in comparison to the original and statistically discretized dataset.