Authors:
Jan Kaniuka
1
;
Jakub Ostrysz
1
;
Maciej Groszyk
1
;
Krzysztof Bieniek
1
;
Szymon Cyperski
2
and
Paweł Domański
1
;
2
Affiliations:
1
Warsaw University of Technology, Institute of Control and Computation Engineering, Nowowiejska 15/19, 00-665 Warsaw, Poland
;
2
Control System Software Sp. z o.o., ul. Rzemieślnicza 7, 81-855 Sopot, Poland
Keyword(s):
Cost Estimation, Full Truck Loads, Machine Learning, Regression, kNN, Decision Trees, Gaussian Processes.
Abstract:
Goods shipping supports the operation and the development of the global economy. As there are thousands of logistics companies, there exists a big need for solutions for their daily operation. The shipment can be carried out in many ways. This work focuses on the road transportation in form of the Full Truck Load (FTL). Once the service is supported by the third party, there is a need to have a tool that compares various offers and allows to estimate the cost. Generally, FTLs are used in the long range routes and the estimation of such contracts can be handled in many ways starting from the simple calculators up to data based machine learning solutions. Nonetheless, the need for the cost estimation appears for the short routes, which often support long range ones. Their pricing rules differs from the long range ones and the required approaches should differ as well. This work presents the wide comparison of 35 regression and machine learning approaches applied to the task. The assess
ment is performed using real contract data of several companies operating in Europe.
(More)