A considerable body of regulatory pronouncements attests to the significance of auditing journal entries for ensuring that financial statements are free of material misstatements; however, existing empirical studies have paid insufficient attention to the audit of journal entries. To explore this issue further, this paper proposes a model based on self-organizing map as well as validates this model by performing experiments on a dataset containing journal entries. Empirical results suggest that the proposed model can detect "suspicious" and legitimate transactions with a high degree of accuracy. Further investigations reveal that the performance of the model is robust to varying prior probabilities of "suspicious" journal entries occurring in the population. The findings indicate that the model can assist auditors in detecting "suspicious" journal entries.