Zeolite Synthesis Modelling with Support Vector Machines: A Combinatorial Approach

This work shows the application of support vector machines (SVM) for modelling and prediction of zeolite synthesis, when using the gel molar ratios as model input (synthesis descriptors). Experimental data includes the synthesis results of a multi-level factorial experimental design of the system TEA:SiO2:Na2O:Al2O3:H2O. The few parameters of the SVM model were studied and the fitting performance is compared with the ones obtained with other machine learning models such as neural networks and classification trees. SVM models show very good prediction performances and generalization capacity in zeolite synthesis prediction. They may overcome overfitting problems observed sometimes for neural networks. It is also studied the influence of the type of material descriptors used as model output. Keywords: Support vector machines, machine learning, zeolites, high-throughput synthesis, data mining