Perbandingan Kinerja Support Vector Machine (SVM) dan K-Nearest Neighbor (KNN) dalam Klasifikasi Stunting

Abstract

The evaluation and comparison of the performance of the K-Nearest Neighbor (KNN) and Support Vector Machine (SVM) algorithms are the primary objectives of this study, particularly in classifying stunting conditions among toddlers. The dataset in this study consists of primary data obtained from the Patianrowo Community Health Center in Nganjuk Regency, with an initial total of 1,102 data points on children aged 0–60 months. After data cleaning, missing values were removed, reducing the dataset to 1,067 data points. Subsequently, the data was divided into 853 training data points and 214 test data points using the train-test split method with an 80:20 ratio. The preprocessing stage included removing missing data, transforming labels into numerical form, and normalizing the data using the min-max scaling method to standardize the feature value ranges. To evaluate the model, a confusion matrix was used with the metrics of accuracy, precision, recall, and F1-score. The test results showed that the KNN algorithm with K=5 produced an accuracy of 96.72%, precision of 91.25%, recall of 67.73%, and an F1-score of 77.52%. Meanwhile, the polynomial SVM algorithm demonstrated improved performance with an accuracy of 97.47%, precision of 90.82%, recall of 78.96%, and an F1-score of 82.55%. Based on these results, SVM is considered more effective in classifying stunting in the dataset used.