Brittle fracturing of materials is common in natural and industrial processes over a variety of length scales. Knowledge of individual particle dynamics is vital to obtain deeper insight into the atomistic processes governing crack propagation in such materials, yet it is challenging to obtain these details in experiments. We propose an experimental approach where isotropic dilational strain is applied to a densely packed monolayer of attractive colloidal microspheres, resulting in fracture. Using brightfield microscopy and particle tracking, we examine the microstructural evolution of the monolayer during fracturing. Furthermore, we propose and test a parameter termed Weakness that estimates the likelihood for particles to be on a crack line, based on a quantified representation of the microstructure in combination with a machine learning algorithm. Regions that are more prone to fracture exhibit an increased Weakness value, however the exact location of a crack depends on the nucleation site, which cannot be predicted a priori. An analysis of the microstructural features that most contribute to increased Weakness values suggests that local density is more important than orientational order. Our methodology and results provide a basis for further research on microscopic processes during the fracturing process.