<p>In a developing nation like India, the maintenance of highway pavements to uphold optimal serviceability presents a formidable challenge to engineers, demanding meticulous planning and advanced technical expertise. Pavement prediction models are indispensable for their ability to provide precise prognostications of infrastructure performance. The parameters influencing pavement deterioration exhibit significant variability across different roads within the same network. Hence in this study, K-means algorithm was deployed to segregate the pavement sections into four homogeneous clusters, thereby facilitating more precise and robust modeling. Data on road inventory, pavement condition, maintenance history, traffic volume and surface roughness within the study area were periodically collected over six cycles. The Pavement Condition Index (PCI) was ascertained from data obtained via visual assessment of the type, severity and extent of pavement distress. This research seeks to develop deterioration models for selected flexible pavements along the National Highways in India using Multiple Linear Regression (MLR) and Artificial Neural Network (ANN). Pavement performance prediction models were generated for clustered and non-clustered data, with clustered models showing lower predictive errors. The findings revealed that both ANN and MLR models were proficient in predicting the PCI with substantial accuracy; nonetheless, the ANN models exhibited enhanced accuracy and efficiency. The study confirms the variability in pavement deterioration and highlights the crucial role of clustering sections for efficient pavement maintenance management.</p>