MySQL based selection of appropriate indexing technique in hospital system using multiclass SVM

Abstract

This paper deals with selection of appropriate indexing technique applied on MySQL Database for a health care system and related performance issues using multiclass support vector machine (SVM). The patient database is generally huge and contains lot of variations. For the quick search or fast retrieval of the desired information from the database, it becomes exigent to select and implement the appropriate indexing technique. Multiclass SVM is proposed to be the optimal solution as SVM could be trained with patient datasets to select the appropriate indexing method in the database of MySQL. B-Tree index is directly applied in most of the cases but other indexing techniques such as Bitmap and Hash are also used as per requirement. Using SVM, corresponding to the given search parameter for information, retrieval of information from the patient database results in quick retrieval of required information in the minimum processing time. Various SVM based methods like one against one, one against all, and fuzzy decision function are implemented for classification on standard electrocardiogram (ECG) datasets chosen from the University of California at Irvine (UCI) Cardiac Arrhythmias database. Suitable SVM method to be used is highlighted in the result section. Depending upon different feature values of ECG dataset, the SVM is trained to identify which categories or classes the given data points belong to. In the similar way appropriate indexing techniques will be selected in the case of health care system application by using multiclass SVM for fast retrieval of patient data. If the findings are made an integral part of the hospital system it could facilitate quick and easy retrieval of patient information, doctor information, hospital information etc. from any hospital as per the requirement. Health care system applications will be developed as front end in .Net whereas, back end as database in MySQL.

International Journal of Engineering, Science and Technology, Vol. 2, No. 6, 2010, pp. 119-130