چكيده به لاتين
In this research, the issue of collecting laboratory samples has been addressed. In the clinical specimen collection process, it is essential that all specimens be tested on the day of collection, which is inconsistent with the donated blood collection process and storage is not permitted. Therefore, the main challenge in the clinical sample collection process is to deliver the collected samples to the central laboratory for testing within a specified time after collection. Specifically, this problem can be defined as open vehicle routing problem with capacity and time constraints. It should be noted that there is no cost to build or close the hub in this issue, so it has a high degree of flexibility and dynamism. In this research, a bi-objective mixed integer programming model has been developed. The objective functions considered for this problem are minimization of sample transfer costs and minimization of carbon dioxide emissions. The goal programming method is used to convert two objective functions into a single objective function. In this study, it is assumed that travel time, travel cost, carbon dioxide conversion factor and the demand of laboratories are uncertain. For this purpose, the Bertsimas and Sim robust approach has been used to deal with the uncertainty of the parameters of travel time, travel cost, and carbon dioxide emission conversion factor. On the other hand, due to the availability of historical data on the demand of laboratories, 4 supervised learning algorithms, which is one of the main branches of machine learning science, were used to predict the amount of demand in the Python software environment. Finally, decision tree and random forest algorithms with the highest accuracy of about 96% had the highest accuracy among the algorithms. After predicting the demand of laboratories for each given shift and working day, the mathematical model of the problem is coded in the GAMS software environment and solved with the help of CPLEX solver. By Using goal programming method different weight combinations of objectives were used and a set of Pareto optimal solutions were obtained. This set of Nondominated solutions were provided to the managers of the network of medical diagnostic laboratories to choose from. The weighted compositions selected by network administrators for the first and second objective functions were 0.5 and 0.5, respectively. Finally, by considering this weight combination in the objective function, the mathematical model was validated and by changing the amount of protection level parameters for uncertain parameters, the sensitivity analysis of the results was performed.
