Maintenance of railway infrastructure plays a vital and crucial role in ensuring safety and comfort during travel. The quality of railway track is determined by the Track Quality Index (TQI) which serves as the primary indicator in assessing the geometric condition of railways and forms the basis for railway infrastructure maintenance planning. This study uses the Systematic Literature Review (SLR) method to comprehensively review and analyze the development of TQI prediction methods. Through a systematic literature selection process from the Google Scholar, Scopus, MDPI and Indonesian national journals databases covering the period of 2010–2025, six studies relevant to this topic were identified. The six studies encompass statistical approaches, machine learning, and Bayesian Network for predicting the TQI values. The analysis result indicated that linear and stepwise regression methods serve as initial approach in TQI modeling, but exhibit limitations in capturing nonlinear relationships between geometric parameters. Further advancements have been achieved through machine learning such as Principal Component Analysis (PCA) and Tree-Augmented Naive Bayes (TAN-BN) which have improved prediction accuracy using railway big data. Recent global trends demonstrate a shift toward Bayesian-based probabilistic models to address uncertainty and enhance reliability in track condition assessment. A major limitation of previous studies is the lack of integration of spatial and temporal data in prediction models. Future research should focus on integration of Bayesian models with deep learning for real-time TQI prediction based on the Internet of Things (IoT).
Copyrights © 2026