Research On Outlier Detection Algorithms For Intelligent Attendance Applications

With the arrival of the information age,daily attendance management can no longer rely on purely manual supervision for enterprises and universities.Using computers to perform attendance management not only saves time and costs to a great extent,but also enables a more comprehensive oversight function.At the same time,it has achieved certain results in improving attendance.Attendance methods are increasingly diverse and intelligent,the ever-changing changes in smartphones have allowed more people to turn their research objects into mobile clients.Based on the service mode of Web App,this dissertation designs an attendance system based on Baidu map API and outlier detection,it mainly uses the mobile positioning function of the smartphones to determine the location of the attendance personnel,and obtains the Mac address of the phone to prevent the occurrence of the substitutive signature,finally,it stores the attendance information in the database for encryption to prevent the disclosure of information.Based on the actual project requirements,this system is mainly used in the university classroom environment.The main work includes the following aspects:(1)Research on one-dimensional outlier detection algorithm.The traditional one-dimensional outlier detection algorithm was analyzed and researched.The robust quartile detection method is selected as a one-dimensional outlier detection method due to robustness requirements and distribution type latitude requirements.In view of the problem that the robust quartile detection algorithm is affected by the middle span of the data,the effect is eliminated as much as possible by extremal filtering of the data columns,and an improved reciprocal difference filtering based on robust quartile detection is obtained algorithm.(2)Research on two-dimensional outlier detection algorithms.In order to avoid the difference in the contribution of the coordinates to the distance values,the Mahalanobis distance is used instead of the traditional Euclidean distance.Based on the one-dimensional outlier detection algorithm,a robust Mahalanobis distance two-dimensional outlier detection algorithm for comprehensive distance and statistical detection is proposed to detect two-dimensional outlier data.(3)Based on the above algorithm design,a smart time and attendance system was developed and applied in the current university classroom environment.The system uses the Baidu map API to obtain student's geographic location information and simultaneously obtains the Mac address of the mobile device to prevent generation of signatures,and uses the DES encryption algorithm to store student attendance latitude and longitude data into the database.Finally,the validity and practicability of the algorithm are verified through experiments.