Technical job distribution at BSD SHARP service center using combination of naïve Bayes and K-Nearest neighbour

• Main Authors: Pebrianti, Dwi, Ariawan, Angga, Bayuaji, Luhur, Mahdiana, Deni, ,, Rusdah
• Format: Conference or Workshop Item
• Language: English; English
• Published: IEEE Computer Society 2022
• Subjects: T Technology (General); T10.5 Communication of technical information
• Online Access: http://irep.iium.edu.my/101597/1/Hybrid_Method_for_Churn_Prediction_Model_in_The_Case_of_Telecommunication_Companies.pdf; http://irep.iium.edu.my/101597/7/Scopus%20-%20Document%20details%20-%20Technical%20Job%20Distribution.pdf; http://irep.iium.edu.my/101597/; https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=9942772

Works distribution is a routine carried out every day by the head of the branch in the SHARP Service Center. The accuracy of the labor division is very important to get customer satisfaction. Inappropriate work distribution can increase complaints from customers. Currently, works distribution in SHARP Service Center is carried out manually, where the works received on the selected system is then shared through the document provided. Time taken for this process is about 1.42 minutes on average for each damage reports. Speed of Service also depends on the Head of Department's expertise and experience. In this study, an automatic system based on Machine Learning will be designed for the technicians work distribution by using a combination of k Nearest Neighbor (k-NN) and Naïve Bayes. Naïve Bayes algorithm is used to improve the feature extraction accuracy by considering the feature below the average (α). Meanwhile, k-NN algorithm is used to classify the experimental data. From the study, it is found that the best of k value for k-NN algorithm is 15. It is known that a high number of accuracy values, the labor distribution can be more accurate. The validation of the proposed method is conducted by using a confusion matrix with a composition of 80% training data and 20% test data. The single Classifier test with the Naïve Bayes algorithm produces the highest accuracy value of 72.7%, while using k-NN algorithm is 81.5%. With a combination of Naive Bayes and k-NN algorithms, the accuracy value is increasing to 86%. This result shows that the proposed method improves the accuracy by 13.3% on single Naive Bayes algorithms and 4% on a single k-NN algorithm. The results obtained show that in the manual process, the average time per job is 1.42 minutes, while by using the proposed method, the average processing time is around 0.03 seconds per job. An increase of 2480 times faster is found and confirmed during the implementation of the proposed method.