Optimasi Algoritma Random Forest dengan PSO untuk Peningkatan Akurasi Klasifikasi Kampanye Digital

Abstract

Digital marketing campaigns are becoming increasingly complex with large and diverse data volumes. This research proposes an innovative method for classifying digital marketing campaign success by optimizing the Random Forest model with Particle Swarm Optimization (PSO). The purpose of this study to improve prediction accuracy and reduce overfitting in campaign data analysis. Using a digital marketing dataset consisting of 8000 data points and 20 variables, the research compares the performance of standard Random Forest with PSO-optimized Random Forest. Results show that the PSO + Random Forest approach consistently provides performance improvement, with the highest accuracy reaching 90.19%, precision of 89.36%, and F1 Score of 88% in the 60%:40% data split. This method effectively reduces prediction errors and produces a more stable classification model. The research contributes to the development of digital marketing data analysis methods and demonstrates the potential of machine learning algorithm optimization for more accurate data-driven decision-making.