Outlier detection for geodetic nets using ADALINE learning algorithm

Abstract

Developed by imitating the operation of human brain, artificial neural network applications are used in many fields such as engineering, industry, medicine, agriculture, finance, communication, meteorology, space and aeronautics. By the help of sophisticated computing technologies, the learning algorithms used in artificial neural networks allowed solving many problems that remained as undecided and defied any mathematical expression, particularly in the fields of engineering. In geodetic studies, three-dimensional geodetic networks are used for all sorts of location-based engineering measurements on earth. Numerous measurements are performed to determine the position of the points in geodetic networks. Possible errors and inconsistencies in these measurements affect geodetic network precision. Therefore, the test for outliers is implemented to eliminate measurement errors and sort out outliers. In the present study, the test for outliers was performed on a computer program developed by using ADALINE learning algorithm and the results were compared with traditional methods (data snooping, Tau, t). This new method was observed to be superior to traditional methods with regards to calculations about outliers and decision-making on the results. Key words: Outliers, neural networks, ADALINE learning algorithm, geodetic nets.