Accident Prediction Model Using Machine Learning. Accuracy of Predicted Model

Abstract

Logistic regression is a predictive model machine learning algorithm that displays the results in a binary form, mostly used in prediction multivariable and as an advanced version of linear regression, here we used to predict the accuracy of our model. Current accident prediction model in Norway is Tusi is a risk based model used in predicting tunnel accidents but it has some problems and loose ends that could make prediction go wrong with minor deviations like human errors, availability of limited data or no data from similar tunnels and is more dependent on previous statics. Past research and this paper have shown the need for the application of ML in accident prediction in road tunnels, the study aims towards how the machine learning models can help in overcoming the problems and replace the traditional method. For the implementation of machine learning algorithm the daily traffic data and accident data is gathered from Norwegian public road administration for one of the existing tunnels of the West Norway known as Eiksund tunnel for examining the correlations in the available variables. Analysis of testing results demonstrated that why the machine learning model needs more study and expertise in computer language, also the calculation results shown the narrow gap of study could have been considered. It is recommended to have a better understanding of the algorithm since the logistic regression is a powerful tool for predictive analysis and there is also a need for better data gathering before the model is trained to strengthen the outcomes.