ABSTRACT
In this chapter, an analysis based on unsupervised learning is presented to find associations and uncovered structures from data of road closures in Colombia. This study was necessary in response to the major logistical challenges facing the country, which combines limited infrastructure with frequent interruptions and high transportation costs. This chapter presents a descriptive analysis of the behavior of road closures in Colombia during the year 2018 and the first quarter of 2019. To this effect, two models of unsupervised learning were applied: clustering k-means and FP-growth. The main variables addressed were the closures of roads and their relationship with the risk factor and time to recovery post-disruption. Afterward, we illustrate the approach on the main road transportation network in Colombia and the transportation of commodities of its six most representative economic sectors, to understand the potential effects on the related supply chains. Around 29,127 events were analyzed for the 32 Colombian departments. Among others, we found a high correspondence between the partial closures of the roads and two predominant causes: landslides and maintenance of the road. Both were also found to be highly influential in affecting the flow of commodities between major cities.
