Shape memory polymers generated by digital light processing are getting attention owing to its process-inherent high resolution and fast curing speed. However, in comparison to shape memory alloys, shape memory polymers show low recovery cycles and consequently are not apt for industrial applications. In addition, for application to the body, the glass temperature of the shape memory polymer should be lower. Previous research using a tBA-co-DEGDA multicomponent photo-resin has shown promising results of 20+ shape memory cycles. However, material characterization in their work has not covered all possible combinations of the individual photo resin components and the glass temperature is high (65°C). To overcome the high glass temperature, we propose the implementation of machine learning for material characterization. To overcome the high volume of data used to train machine learning algorithms, design of experiments and machine learning model are implemented. Design of experiments is used to obtain uncorrelated data with minimum experiments whether as machine learning is used to identify relationship between input and output. Through this manner, relations between input (material compositions and printing parameters) and output can be determined. In this case, output is defined as the glass temperature. Through design of experiments and machine learning as well as optimization, optimal resin compositions are proposed.