Classification of surface condition of flexible road pavement using Naïve Bayes theorem

Abstract

Data collection, availability and accuracy on road pavement surface condition have been identified as the major setback for pavement management system in low and medium income countries. This situation necessitates the adoption of historic dataset which often is characterised by missing and noisy data elements. The Naïve Bayes theorem which depends on conditional probabilities of prior events or attributes for condition classification and prediction was used to investigate the authenticity of surface condition classification of flexible road pavement. Some selected links of Federal Highways in Nigeria based on the availability of historic dataset were considered. The Naïve Bayes theorem as a data mining approach was implemented using Waikato Environment for Knowledge Analysis (WEKA) software which performed satisfactorily in surface condition classification with minimal margin of errors due to its ability to handle challenges of missing and noisy dataset. Results of classifications indicated that; links 716, 5, 8, 22 and 136 in Borno, Kwara, Lagos, Oyo and Plateau states respectively had relatively high level of unworthiness as at the time the data was collected, hence call for immediate maintenance and rehabilitation actions, while links 89, 375, 370, 130, 17, 332, 138, 144 and 255 from Anambra, Bendel, Imo, Kaduna, Ogun, Plateau, Rivers and Sokoto states respectively were worthy and had no cause for immediate maintenance.