Predicting the Effectiveness of Preeclampsia Medications Based on Dose and Method of drug Consumption Using Data Mining

Preeclampsia is a serious disorder during pregnancy that even can lead to termination of pregnancy and fetus abortion. This study was performed with the aim to generate an efficient predictive model to accurately predict the effectiveness of preeclampsia medications. This is a descriptive, cross-sectional study performed on 774 patients in Neyshabour Hakim Hospital from 2011-2013. Three algorithms, C5.0, C & R Tree and CHAID were applied on data of preeclampsia, and then the accuracy of generated models was obtained. In this study, Clementine software (version 12.0) was used for data analysis and implementation of data mining algorithms. The obtained values for accuracy of generated models by implementation of C5.0, C&R Tree and CHAID algorithms on train and test dataset were 99.63 and 99.14, respectively. The high accuracy of models showed good performance of these algorithms. The sensitivity of 100% and precision of 99.5% confirmed that C&R Tree is superior to other algorithms. Obtained results of clustering on better tree showed that when the effectiveness of medicine has been increased, the patient's mean age was increased, and it was due to direct relation between preeclampsia and increasing age. The obtained results show that the probability of preeclampsia is increased with increasing age and the medications with high effectiveness should be used for treatment. Therefore, more attention should be taken for selecting the used medicine with different effectiveness. Prediction of the effectiveness of medicine is performed for achieving these targets: help to the doctor for increasing the accuracy of diagnosis and prevention of incorrect diagnosis in dose of medicine consumption for patient, diagnosing the preeclampsia severity and inhibition of dangerous effects of extra consumption of medication by patient, prediction of the adequate medicine storage and prevention of medicine lack side-effects.