<![CDATA[In this study, we compared the performance of the K-Nearest Neighbor (KNN) and Support Vector Machine (SVM) algorithms for predicting pregnancy risk in patients. The dataset used consisted of eight variables, such as number of pregnancies, vaccination status, blood pressure, fetal heart rate, body mass index (BMI), age, and height, which were used to classify pregnancies as high-risk or low-risk. Data preprocessing was performed by standardizing numerical features and dividing the data using a stratified split method (80% training data, 20% test data). Model evaluation results showed that KNN achieved an accuracy rate of 81%, while SVM achieved an accuracy of 75.50%. Further analysis showed that KNN was more stable in classifying data with diverse variable distributions, while SVM showed a tendency to be more sensitive to high-risk cases, with 66.7% of predictions pointing to that category. Compared with previous studies, KNN's performance was within the general accuracy range of 70–95%, while SVM's accuracy was slightly lower than the average of similar studies, which reached around 80.33%. This study concludes that both algorithms have the potential to be tools for early detection of maternal health risks, where the differences in their performance are influenced by the selected parameters, data characteristics, and class division.]]>
Copyrights © 2025
