The transformation of actual processes into sustainable processes is a major study subject over recent years, particularly through the circular economy. However, the environmental assessments require a huge quantity of data and many of these data are heterogeneous. Environmental evaluation tools would clearly benefit from Data Science approaches in the Big Data context. This paper focuses on developing a framework for decision-making in Process System Engineering by coupling Machine Learning techniques and environmental assessment. Five-steps framework have been deployed in a framework and tested on the comparison of biomass pretreatment processes for glucose production. Some scientific articles have been selected thanks to specific keywords in Science Direct and Web of Science. The data architecture and in particular the data analysis allows us to bring data to higher quality such as a material balance check. The approach gives access to a process-impact matrix which is analyzed through Dimensional Reduction methods in order to highlight similar impacts and/or processes.