Industry cluster identification has become an important research topic in regional science, partly as a response to the growing demand by policymakers for analytical tools that provide a better understanding of a regional economy. The main objective of this paper is to provide a detailed description of a new approach to identify industry clusters and interindustry networks, based on input-output tables. The goal is to outline and efficient algorithm using readily available data so that the method can be replicated in any region, thereby allowing for comparative studies of industrial clusters over space and time. This new method draws on concepts from network analysis theory. It can be divided into two main block: data reduction and network partitioning. Data reduction begins by representing industries and products/services flows as a directed graph, where the links are represented as arrows indicating the direction of the flows. Based on several assumptions about how industries are related into a supply chain, the initial graph is then concerted into an undirected graph by transforming products/services flows into relative weights. The second reduction is formulated as a minimization problem, resulting in a minimum spanning tree (MST) for a subset of the initial graph edges. The final clusters are obtained by selectively deleting edges in the MST, such that each cluster contains a core industry.