Feature Weighting Optimization: Genetic Algorithms and Random Forest for Classification of Pregnant Potential Risk

Many pregnant women still do not want to check their pregnancy to recognize the signs of high-risk pregnant women. High-risk pregnancies refer to abundant women with several risky conditions requiring special care during and after delivery. This study aims to predict pregnancy risk levels with early detection efforts. The method used in this research is the Random Forest method for classification and the Genetic Algorithm for optimization. The results obtained from the Maternal Health Risk dataset analysis using the Random Forest method get an accuracy value of 73.37%. After optimizing the features using a genetic algorithm, the accuracy value in the classification of the health risk level of pregnant women has increased significantly to 90.20%. The results of the t-test in this study showed a value of less than 0.05, and the t-test value obtained is 0.000218269. That is, the genetic algorithm has excellent performance to improve the performance of the classification algorithm. Based on the results of the study to get a good accuracy value, it can be used by medical personnel to detect the level of health risk in pregnant women. It can be a reference material for further research to implement this research into a program.