Research 1 (Working)

A Novel Online Fair Task Allocation Using Intelligent Randomized Assignment: Machine Learning Approach:

This article investigates the problem of fair task allocation in an electronic devices service-repair company. The variety of repair operations, as well as the difference in the cost of repairing each of them, was the main motivation for designing a fair method for allocating repair devices (task) to the technicians of this company. Four different allocation scenarios were designed and evaluated to achieve this goal based on the amount of information at hand. To design these allocation scenarios, which are completely online, Integer programming (IP), Machine Learning (ML) approach, as well as random allocation and their combination approaches were used and compared. Also, in one of the scenarios, an intelligent random allocation was designed by combining the random allocation method and machine learning, which performed the best for the case study. Also, to design this scenario, a Hierarchical Ensemble Classifier by utilizing KNN and SVM was used to predict the repair revenue of repair devices, which had an accuracy of 65% for prediction. At the end, by assuming the existence of 5 technicians and based on the company's real data, performances were evaluated in Python and GAMS, and the unfairness index of different allocation scenarios was studied and compared, which led to selecting the best allocation scenario for the problem.

The SQL Server report of the database of the company:

The Machine Learning methodology of the paper:

Plot of the invented hierarchical ensemble classifier:

The plot of the performance of  Machine Learning based allocation scenario for each month:

The plot of comparing the performance of all allocation scenarios of this article:

The plot of comparing the accuracy of the devised predictive model with human performance: