The virtual representation of physical objects, processes or service, that either exists currently or that is intended to be developed is called Digital Twin. It can be a replica of an object as small as a switch to a giant thing like a windmill or can represent a geographic region. Most of the time the developed replicas are used to study a process under controlled settings. In other words, it is a kind of simulation of the object or the process meant for collecting data, being fully present in a form similar to the original product or service.  Predictive analytics is a data heavy process and so the digital twins are the staple for new models. Given the quality and prompt data a digital twin can provide, making a real-time connection is way easy. Therefore, predictive analytics and digital twins are closely related and dependent on one another. There are two ways one can understand this relationship. 

 Firstly, As explained earlier predictive analytics uses data generated by digital twins. Secondly, the very process the data is collected. In a real-world scenario, the data is generated against a large number of variables, but what the users might want is a simpler model with limited variables. A digital twin is a more precise and complicated version of a predictive model. Therefore, digital twins carry the benefit of testing the models for fewer variables, all while helping in looking at the big picture.