Patents have long been recognized as a rich source of data for studying innovation, technical changes, and value creation. Patent data includes citations to previous patents, and patent citations allow one to create an indicator of patent value. Identifying valuable patents in a timely manner is essential for effectively harnessing the business value of inventions in the increasingly competitive global market. However, the existing methods of evaluating patent value suffer the issues of timeliness and accuracy. In this paper, we propose a data mining approach that utilizes the structural properties of patent citations networks to predict the value of patents while aiming to improve timeliness and accuracy.