Machine Learning–Based Recommendation System for Optical Distribution Point Placement in Fiber Access Networks

Abstract

The rapid expansion of the digital economy requires reliable telecommunication infrastructure, particularly fiber optic access networks that provide high-speed broadband connectivity. One critical component in these networks is the Optical Distribution Point, which functions as a distribution node connecting optical infrastructure to end users. However, ODP placement is often determined manually, leading to inefficient resource utilization and inconsistent decision-making. This study aims to develop a data-driven recommendation system for optimal ODP placement. The proposed approach integrates spatial feature engineering with supervised machine learning techniques to analyze infrastructure capacity, spatial distance, and customer distribution. Several algorithms were evaluated, including Random Forest, Logistic Regression, K-Nearest Neighbors, Gradient Boosting, and a Stacking Ensemble model, while Synthetic Minority Oversampling Technique was applied to address class imbalance. Model performance was evaluated using Precision, Recall, F1-score, ROC-AUC, and Normalized Discounted Cumulative Gain. The results show that Gradient Boosting achieved the highest performance with an F1-score of 0.8986 and ROC-AUC of 0.96, while the Stacking Ensemble model demonstrated stable ranking performance with a mean NDCG of 91.75%. The proposed system improves the efficiency and accuracy of ODP placement planning and supports data-driven telecommunication infrastructure development.