As the development of new composite materials can take over 20 years, a less time-consuming approach to considering the vast number of possible combinations of materials and predicting their physical characteristics is required. Material informatics, a field of study combining information technology and material science, is one such promising approach whereby researchers use information technology to predict the physical characteristics of new composites. In the present study, we adopted a link prediction approach of co-occurrence words to detect possible combinations of previously uncombined materials in order to assist the development of new composite materials. We observed research areas of composite materials and estimated their growth in the immediate future. Using the bibliographic data of academic papers, we extracted co-words from the titles and abstracts and created co-occurrence networks of material words. From these networks, we extracted the optimal materials for detecting new combinations. This research contributes to the acceleration of the design of new composite materials by reducing the time and cost to research and develop new materials.