Classification of Family Planning Participant Status Using Support Vector Machine (SVM) Based on Age and Type of Contraceptive

Abstract

This research investigates the application of the Support Vector Machine (SVM) algorithm in classifying the status of Family Planning (FP) participants based on two primary input features: the wife’s age and the type of contraceptive method used. The classification objective includes determining whether a participant falls under the categories of “New,” “Repeater,” or “Dropout.” The dataset used in this study was sourced from the Population and Family Planning Control Service Unit (SATPEL PPKB) of Cilebar District, Karawang Regency. It consists of 1,402 records containing both demographic information and contraceptive usage data. Initial data preprocessing involved label encoding for categorical variables to make the dataset suitable for machine learning algorithms, followed by an 80-20 split for training and testing purposes.

The SVM model was trained using the processed dataset, and its performance was evaluated using standard classification metrics such as accuracy, precision, recall, and F1-score. The model achieved a classification accuracy of 56.2%. While it demonstrated reasonable effectiveness in identifying the majority class—“New” participants—with relatively high precision and recall, the model's performance declined substantially for minority classes, particularly the “Dropout” group, which was not accurately predicted by the model. This significant discrepancy in classification performance is likely attributed to class imbalance and the limited discriminatory power of using only two input variables.

The results indicate that, although SVM is capable of modeling the problem, its performance is constrained by the simplicity of the input features. The model's inability to generalize across all classes equally suggests that additional predictive features—such as education level, number of children, duration of contraceptive use, and socio-economic status—may be necessary to improve overall classification accuracy. Furthermore, addressing the imbalance among class distributions through resampling techniques or cost-sensitive learning may enhance the model’s capacity to recognize underrepresented groups. In conclusion, this study serves as a foundational attempt at FP status classification using SVM and highlights the importance of comprehensive feature selection and data balancing strategies for more effective future implementations.