Özet:

Abstract One of the most prevalent metabolic conditions that raise blood sugar is diabetes. The complexity of this condition's interdependence on different circumstances makes early detection difficult. To aid medical professionals in the diagnosing process, crucial decision support systems must be developed. This study suggests creating a type 2 diabetes predictive model with high classification accuracy. The study was divided into two main categories. The proposed data mining model can be viewed as to consist of distinct phases namely classification and clustering. Two Classification phase includes the analysis of Data Mining Classification Algorithms and identifying the best algorithm for the health dataset. Classification is performed on Indian Pima Diabetes Dataset. The clustering framework is a method to identify groups within patient records and to detect clusters of attributes that can be used to identify the target class. This is the proposed data mining model that can detect clusters and classes in a given patient record. For achieving better accuracy, clusterization is performed on second dataset i.e., Indian Patient Liver Dataset. The study suggests adopting a bigger population sample without geographic restrictions as a potential solution. Also, it would be interesting to investigate tuning on top of normalised data to further improve accuracy because the created SVM model did not undergo any special hyperparameter tweaking.

Anahtar Kelimeler: