K-Nearest Neighbors Optimization using Particle Swarm Optimization in Selection Digital Payments

Abstract

Fintech developments have increased the use of digital payment systems such as OVO and GoPay. However, selecting a payment method that suits user preferences is still a challenge. This research proposes a combination of K-Nearest Neighbors (KNN) and Particle Swarm Optimization (PSO) to improve the classification accuracy of digital payment systems. The dataset used comes from a survey of Fintech users with factors such as ease of application, data security, cashback and customer service. KNN is used as a classification method, while PSO is applied for feature selection to improve model efficiency. Evaluation is carried out using accuracy, precision, recall, and AUC. The research results show that accuracy increased from 94.00% to 95.47% after optimization with PSO. The most influential factors are customer service, user employment and cashback. However, the AUC value remains 0.500, which shows that the model still has limitations in optimally differentiating categories. Further research is recommended to explore other algorithms such as Random Forest and SVM, as well as developing a machine learning-based digital payment recommendation system