Predicting Dielectric Polymer Component from Different Properties via Machine Learning Method

Designing linear polymer of excellent insulating properties is always a disturbing problem in polymer design aspect, especially positively design which means positively get structure formula from possible properties. Many inspiring methods have been put effort into the aspect. In this paper, a new method has been proposed for predicting possible fragments in linear polymer and design the possible structure from targeted properties. Random forest method has been used for establishing the connection between five properties and 362 possible seed fragments. In the end, an example with properties values of PEI has been put through the method and different candidates with similar properties has been assembled including PEI polymer.