Optimizing KNN for Face Recognition and Location Detection in Mobile Employee Attendance Systems Using Machine Learning

Abstract

Employee attendance is critical to company operations, particularly for firms involved in ATM installations collaborating with banks. An effective attendance system enables bank vendors to verify the completion of ATM installations. The current attendance method employed by the company requires employees to send photos of themselves with the installed ATM, including the machine number and installation location. The company's admin then compiles these photos into a summary that serves as proof of employee attendance and job completion, which is subsequently sent to the bank vendor. However, this method fails to accurately verify the employees' physical presence at the installation sites, raising significant concerns about the validity of attendance data. This lack of accuracy in attendance data affects performance evaluations, vendor trust and satisfaction, and the company's operational integrity. A facial recognition attendance system using the K-Nearest Neighbors (K-NN) method was proposed to address these issues. The K-NN method was selected for its effectiveness in classifying data based on proximity characteristics, making it ideal for precise attendance verification based on location and time. Test results demonstrated that the system achieved a 100% success rate in distance detection, 96.4% in position detection, and 50% in accessory usage detection. This system enhances the efficiency and accuracy of the employee attendance process.