Development of Ice Prediction Mobile Application

Cold weather and heavy winter conditions cause icing on the roads, and therefore many fatal, injured and materially damaged accidents occur every year. In this study, an icing prediction algorithm and mobile application has been developed to prevent accidents caused by icing on the roads. With the developed application, it is aimed to give preliminary information about the formation of icing in line with the routes of the drivers. In the study, the temperature, dew point, sensed temperature, wind intensity, wind direction, relative humidity, wind speed input parameters were taken from the road condition sensor and weather stations. At the exit, double classification was made with icing information. After the training of the system is completed, weather forecast information is obtained from the meteorology and icing forecast is made for the next 12 hours on the developed mobile application. In addition, in order to measure and compare the accuracy of the developed system, the multi-layer perceptron (MLP) neural network model and linear and nonlinear support vector machines (SVM) methods are used. Considering the classification accuracy of the algorithms used in the study, based on the total number of correctly classified samples, it was seen that the model of the MLP performed best with 87,26% accuracy rate, followed by the linear SVM model with 86,32% and our proposed model with 75,47% accuracy rate. However, in the developed prediction algorithm, although the classification accuracy is lower compared to others, it has been observed that the number of samples used in training increases, the icing prediction accuracy increases in direct proportion.