Using Information Gain To Evaluate Weigh-in-motion Axle Load Management Information System

Abstract

Delays and other operational inefficiencies at weighbridges in Kenya has remained persistent over the years. This delays have led to astronomical rise in transport costs to transporters. These costs are mostly passed on to consumers of the transported goods. Causes of the delays at weighbridges stem from inspection of legal requirements such as axle load limits, validity of load permits, validity of driving license, drunk drivers and possession of fire extinguisher equipment. Not all trucks screened at the weighbridge violate these laws and regulations. The research therefore sought to establish a pattern using data mining techniques and tools from data generated from the weighbridges. Data on origins, destinations and compliance o these requirements was obtained from weighbridge data at Webuye station. Out of tens of thousands of data at the database, two thousand five hundred and fifty data items were randomly selected for this study. The selected sample was divided into training data set for building the model and testing data set for testing the model. WEKA data mining software was used to perform the data mining. J 48 classification algorithm in WEKA was employed to build the model by establishing a patter that linked origins, destinations and the likelihood of committing the offences. Performance metrics of the model indicated that there was a strong predictive accuracy by the model using the F-measure 79.9%, True Positive Rate 82.7%, Recall 82.7% and Precision at 81.1%. from the results of this study decision makers and policy formulators at the weighbridges may use them to institute policy that will address vehicle screening process based on their destinations and origins with an aim of implementing operational efficiency at the weighbridges.