Methodology to explore open data of road crashes using Data Science: Case Medellín

Abstract

Road crashes is a serious public health problem in the world. The study of open data on this subject
can stimulate more timely and informed decisions. The objective is to propose a methodology to study
open data on road accident (Medellín case) using Data Science, considering from the planning of the
study to the web visualization. The methodology consists of four macroprocesses: 1. Planning, 2. Data
preparation, 3. Automatic analysis and 4. Data visualization (web application). These consist of one
or more stages, disaggregated into 15 sub-stages with univariate, bivariate and multivariate scopes.
Macroprocesses 2-4 were automated in R language. As a result, the analyst can become familiar with the
topic (descriptive), explore relationships between variables, locate events, induce patterns of grouping
and identify some factors associated with the events of accidents. All these, combining variables for a
more detailed segmentation. The case study also has value for other areas, since road accidents generate
greater effects in developing countries, which is attracting the interest of researchers.