Clustering and Classification Analysis in Financial Reporting of Philippine Government Business Enterprises

The use of machine learning to detect unusual or anomalous financial reporting in government financial statements is generally overlooked in the literature despite the possibility of fraudulent financial reporting occurring not only in private organizations but also in the public or government sector. In this paper, clustering and classification algorithms were applied to analyze financial reporting patterns, identify unusual observations, and develop an anomaly detection model in the financial reporting of Philippine Government Business Enterprises. Using financial and non-financial variables relevant to government financial reporting, a clustering model using Density Based Spatial Clustering of Applications with Noise was developed to label observations, whether usual or unusual. The labeled dataset was further used to build a Decision Tree classification model to identify financial statements exhibiting anomalous patterns. The results show that significant variables that may indicate an unusual behavior are classified under fraud indicators, audit findings, and audit recommendations. The detection model developed in this study is an effective decision support tool to government auditors to signal unusual observations in aid to planning the audit accordingly. Likewise, the structure of this study is applicable to financial reports of other government agencies relevant to conduct a comparative study.