<p>Bankruptcy prediction has been a long-standing and highly sought-after field of research for many years. Several researchers have utilized a wide range of techniques, including multivariate discriminant analysis, artificial intelligence (AI), and machine learning-based models, to forecast bankruptcy. Nevertheless, there is a scarcity of studies that employ network-based data for bankruptcy prediction, resulting in a conspicuous void in this area of research. This paper introduces a methodology that uses network analysis to assess the probability of bankruptcy among organizations by examining their similarity in important financial characteristics. Our analysis leverages multiple comprehensive datasets encompassing diverse geographical regions and industry sectors. Each dataset contains extensive financial variables and corresponding bankruptcy status information. We identify the five most influential features in each dataset, to build the network and then produce five matrices that represent the similarity between organizations for each feature. Subsequently, these matrices are combined to create a full similarity matrix. By employing this matrix, we build a network structure between the companies, extract seven crucial network-centric characteristics, employ community detection to group companies into clusters, and assign cluster labels to the dataset features. By including these characteristics into the initial dataset, we utilize machine-learning methods to forecast bankruptcy. In order to assess the accuracy of the model, we compare the predictions generated using the original and enhanced datasets. The results demonstrate consistent performance and robust predictive capability across different economic contexts, market conditions, and industry sectors, showing significant enhancement in prediction accuracy when utilizing features derived from the network.</p>