Novel Approaches for Transport Infrastructure Reduction to Effective Optimisation of Flow Tasks

journal article in Scopus

en

Nowadays, increasing complexity of solved optimisation problems leads to necessity of dealing with computation time demand. In the case of network flow tasks, computation time is highly dependent on detail of transport infrastructure. The presented paper is concerned with developing novel approaches for transport infrastructure reduction using clustering analysis. According to the required outputs of the task, it is possible to variably change the detail of the network in individual territorial units to ensure the solvability of the task, but without significant distortion of the results. The main idea and novelty of the presented research is to have a finer construction only in the vicinity of the monitored subject. With a greater distance, it is possible to reduce the level of detail in the transport network. The principle of reduction technique is based on transformation of geographic coordinates with subsequent cluster analysis. K-means and hierarchical clustering are introduced and results of developed approach are shown on municipalities in Czech Republic. Consistency within clusters of both methods is evaluated using silhouettes. The presented methodology allows to solve optimisation of case studies more efficiently with greater detail in monitored region, which leads to more accurate solutions.

Nowadays, increasing complexity of solved optimisation problems leads to necessity of dealing with computation time demand. In the case of network flow tasks, computation time is highly dependent on detail of transport infrastructure. The presented paper is concerned with developing novel approaches for transport infrastructure reduction using clustering analysis. According to the required outputs of the task, it is possible to variably change the detail of the network in individual territorial units to ensure the solvability of the task, but without significant distortion of the results. The main idea and novelty of the presented research is to have a finer construction only in the vicinity of the monitored subject. With a greater distance, it is possible to reduce the level of detail in the transport network. The principle of reduction technique is based on transformation of geographic coordinates with subsequent cluster analysis. K-means and hierarchical clustering are introduced and results of developed approach are shown on municipalities in Czech Republic. Consistency within clusters of both methods is evaluated using silhouettes. The presented methodology allows to solve optimisation of case studies more efficiently with greater detail in monitored region, which leads to more accurate solutions.

Model-size reduction techniques, Cluster analysis, Infrastructure transformation

15.11.2021

AIDIC S.r.l.

Milano, Italy

2283-9216

88

1

463–468

6