Improving every child's chance in life

This article describes the results of a data mining project designed to explore the key drivers of the Australian Early Development Index (AEDI), a numerical indicator of early childhood development vulnerability. The work was conducted during GovHack 2013, a 48-hour Australian Open Data competition where participants were required to use published open data sets provided by various Australian government and other agencies. We applied advanced machine learning techniques (random forests, generalised boosted regression models and multivariate adaptive regression splines) to the South Australian state and national data to gain insights into the key drivers of AEDI and to quantify the levers that the state government, community and individuals could apply to improve the situation. We found that after accounting for the population specifics and socioeconomic conditions, for example unemployment level and Index of Relative Socioeconomic Disadvantage, the most important factors impacting early childhood development were lack of motor vehicle in the household, inability to afford buying medication and maternal smoking during pregnancy. We quantified the impact of each of these factors and suggested relevant potential Government interventions. We then visualised our findings and created a Web app that allowed various intervention strategies to be interactively explored, based on the derived relationship between early child development index and its key drivers.