The Thesis was focused on developing Machine Learning models that predict catalyst performance with relevant descriptors. Literature dataset was used for modelling selectivity and conversion with two different solvents. A systematic testing of variable selection and modelling methods was performed. The methods are readily usable in WP2 where dehydration reaction of C5 and C6 sugars with simple metal catalysts will be studied. The goal is to identify ML models that will accelerate the catalyst identification with high throughput screening methods, identify non-obvious formulations and allow catalyst tuning for different feedstock compositions.
Link: http://jultika.oulu.fi/Record/nbnfioulu-202101211052
