An Intelligent Decision Support Tool for Supply Perishable Materials: Case Study of Concrete Delivery

Dispatching process of perishable materials must be fast and robust to deal with uncertainties occurred during supply process. Solving large scale dispatching problem with available computing facilities are computationally impractical and it is characterized as NP-hard. In practice and with lack of practical automated solution, experts are hired to handle the dispatching centers and manually solve the resource allocation tasks. This paper introduces a platform to automatically match experts’ decisions by implementing machine learning techniques. As a case study the atuhors selected concrete delivery dispatching problem and tested the developed algorithms with a large scale dataset collected from an active Ready Mixed Concrete (RMC) with 6 active batch plants (depots) and around 100 trucks. The available dataset covers a period of three months and 79 working days. The training dataset is constructed by extracting a set of proposed attributes from a field dataset. Eleven metrics are used for assessing the performance of the selected machine learning algorithm. The results show that machine learning approach can predict the experts’ behaviour with accuracy of around 90% however there is a big concern about the 10% error. To investigate this issue, an algorithm is developed to check the feasibility of misclassified instances obtained by the machine learning. The result of this test shows that around 30% of misclassified solutions are feasible. This shows that the machine learning method can even identify the human errors and ignore them. Finally, the operational costs of experts` decisions and acquired solutions by the machine learning approach are calculated. The results show that despite out-sourcing for infeasible solutions acquired by the machine learning scheme, on average the operation cost of experts` decisions is only 5% lower than the machine learning based solution. The results show that this platform can provide feasible solutions on near 95% instances and therefore it is recommended to be implemented in practice for three main purposes (i) as a decision support system, (ii) as a tool for training new dispatchers (iii) as an alternative for handling RMC dispatching centers by either replacing dispatchers or reducing their size.

• Supplemental Notes:
  • This paper was sponsored by TRB committee ABJ70 Standing Committee on Artificial Intelligence and Advanced Computing Applications.
• Corporate Authors:

  Transportation Research Board

  500 Fifth Street, NW
  Washington, DC  United States  20001
• Authors:
  • Maghrebi, Mojtaba
  • Waller, S Travis
• Conference:
  • Transportation Research Board 96th Annual Meeting
  • Location: Washington DC, United States
  • Date: 2017-1-8 to 2017-1-12
• Date: 2017

Language

• English

Media Info

• Media Type: Digital/other
• Features: Figures; References; Tables;
• Pagination: 19p
• Monograph Title: TRB 96th Annual Meeting Compendium of Papers

Subject/Index Terms

• TRT Terms: Algorithms; Dispatching; Expert systems; Machine learning; Ready mixed concrete
• Subject Areas: Data and Information Technology; Freight Transportation; Operations and Traffic Management;

Filing Info

• Accession Number: 01628209
• Record Type: Publication
• Report/Paper Numbers: 17-06691
• Files: TRIS, TRB, ATRI
• Created Date: Mar 7 2017 10:25AM