Inconsistent transaction records make it difficult for users to understand personal spending patterns. This study implements a Gaussian Mixture Model (GMM) in an Android application to group expense amounts and summarize financial patterns. The application was developed using a prototyping approach with Flutter/Dart and Firebase. The GMM test data comprised 100 simulated expense transactions from August to October 2026, not participant data. Expense amounts were transformed using the natural logarithm and modeled as a univariate three-component GMM estimated by Expectation-Maximization. The model met its convergence criterion at iteration 1,012. Approximate component centers on the nominal scale were IDR 71,721, IDR 419,082, and IDR 2,467,997, with 79, 18, and 3 transactions, respectively. Functional testing recorded 27 passed tests. White-box testing of the financial-capacity assessment flow yielded a cyclomatic complexity of 6, and all six independent paths passed. GMM complements the application’s separate 50/30/20 budgeting rule. Findings are limited to simulated data; cluster quality was not evaluated against external ground truth or internal validation metrics. Results therefore describe implementation behavior and do not establish population-level performance or the validity of the three clusters.
Copyrights © 2026