Pages: 75-85

Abstract

In the present study, the modified XGBoost algorithm is introduced for identifying ‎users across different digital devices. Then, it is implemented on a dataset and the ‎results are compared with the results of the decision tree, support-vector machine and ‎K-nearest neighbor algorithms. Using several experiments with different parameters, it ‎is proved that using 200 leaves and 1000 trees, the best result is achieved, which is the ‎accuracy rate of 99.94% and the corresponding running time is 5439.26 seconds.‎ The running time of the proposed algorithm is much greater as compared to the other ‎algorithms tested in the present study, so, it shows poor performance in running time. ‎But what makes this algorithm significant is its accuracy. According to the results, its ‎accuracy rate is nearly 100% for any number of input data. The proposed algorithm ‎shows that it can provide reliable results in the case of offline running where it does ‎not matter how much its computing speed is‎‎‎.