Data Mining and Performance Prediction of Flexible Road Pavement Using Fuzzy Logic Theory: A Case of Nigeria

Abstract

The over dependence on road transport system to cater for the fast growing human population in some developing countries like Nigeria has necessitated the need for the development of an efficient and sustainable road pavement management system. This study used data mining techniques namely; Random Forest, Decision Tree and Naive Bayes algorithms to examine the inferred dataset on flexible road pavement performance attributes and surface condition classification in Nigeria. The data mining techniques were used to investigate hidden relationship between pavement performance variables and to authenticate the accuracy of subjective measurements that were used for pavement surface condition classification. The Random Forest and Decision Tree algorithms reported perfect classifications of road pavement sections into; Excellent, Good, Fair, Poor and Very poor. On the other hand, the Naïve Bayes algorithm yielded inaccurate classifications with some margin of errors which were attributed to missing and noisy entries in the dataset. This necessitated the use of Fuzzy logic theory for the performance prediction due to its capability to handle the imprecise dataset. It was used to develop Fuzzy Inference System (FIS) for performance prediction of flexible road pavement using attributes such as; the classified Initial Pavement Condition (IPC), Age of pavement, Resilient Modulus (MR) of subgrade soil, Average Truck load per day, Average Annual Air Temperature and Rainfall to predict the Future Pavement Condition (FPC). The model was calibrated using the observed logical behaviour of road pavement to fit the engineering experience and judgement. A goodness-of-fit test between the observed and predicted FPC values showed high level of consistency – correlation coefficient at 90%. The research proposed 5120 mutually exclusive Fuzzy Logic Rules for performance prediction of road pavement based on permutation theory. Though, the required well-spread dataset for calibration of the model to cover all possible pavement conditions in Nigeria and subsequent validation were not available, a framework for performance prediction of flexible road pavement was developed, and a comprehensive guidelines on how to calibrate the FIS model using well-spread dataset was presented.