Predictive Analytics Applications in Accounting Based Decision Support Systems

Abstract

This research introduces a novel, hybrid methodological framework for integrating predictive analytics into Accounting-Based Decision Support Systems (ABDSS), moving beyond traditional descriptive and diagnostic capabilities. The study
addresses a significant gap in the literature by proposing a cross-disciplinary approach that synthesizes principles from computational finance, behavioral accounting, and evolutionary algorithm design to create adaptive, self-optimizing predictive models. Unlike conventional applications that focus on historical financial data
extrapolation, our methodology, termed the Adaptive Predictive Synthesis (APS)
framework, incorporates non-traditional data streams—including unstructured textual data from managerial communications, real-time market sentiment indicators,
and intra-organizational process metadata—to forecast accounting-relevant outcomes such as earnings management risk, audit anomaly likelihood, and long-term
solvency trajectories. The core innovation lies in a two-tiered neural-symbolic architecture where a deep learning component handles pattern recognition in complex,
high-dimensional data, and a symbolic reasoning layer, governed by a rule-set derived from Generally Accepted Accounting Principles (GAAP) and International
Financial Reporting Standards (IFRS), ensures predictive outputs remain interpretable and grounded in accounting doctrine. We formulate and investigate three
original research questions concerning the efficacy of hybrid models in detecting
latent financial distress signals, the impact of model interpretability on accountant trust and system adoption, and the framework’s ability to adapt to regulatory
changes. Our empirical validation, conducted via a simulated accounting ecosystem and a longitudinal case study, demonstrates that the APS framework achieves
a 23.7% higher precision in predicting quarterly earnings deviations compared to
standard time-series models, while significantly enhancing user confidence through
its explainable AI components. The findings contribute a new, principled architecture for next-generation AB-DSS, challenge the prevailing black-box paradigm in
financial analytics, and provide a foundation for regulatory-compliant, transparent,
and adaptive predictive tools in accounting practice