This study examines bankruptcy in terms of financial variables as well as interfirm network structure variables. We first binarize the variables by introducing a threshold and then select the appropriate set of variables that minimize the p-value in Fisher's exact test. Here, the financial variables related to borrowing and savings are strongly correlated with bankruptcy, but the variables of trade network and capital network, including chain bankruptcy effect, have weaker yet significant correlations. Finally, we perform a bankruptcy prediction with the selected variables and a second-order Ising model and confirm that the Ising model has relatively higher predictive power than the logit model.