TITLE:
Applying Deep Learning Techniques for Automated Analysis and Interpretation of Financial Statements
AUTHORS:
Godfrey Wandwi, Christian Mbekomize
KEYWORDS:
Multimodal Learning, Deep Learning, Financial Statement Analysis, LSTM, FinBERT, Financial Text Mining, Automated Interpretation, Financial Analytics
JOURNAL NAME:
Open Journal of Applied Sciences,
Vol.15 No.12,
December
16,
2025
ABSTRACT: The increasing complexity of financial statements, which encompass both structured numerical data and unstructured textual narratives, presents significant challenges for traditional analytic approaches. This study proposes a multimodal deep learning framework that integrates Long Short-Term Memory (LSTM) networks and FinBERT, a domain-specific transformer model pre-trained on financial text, to enable automated analysis and interpretation of financial statements. The architecture is designed to capture temporal dependencies in financial metrics through LSTM while extracting semantic meaning from textual disclosures using FinBERT. By fusing both data modalities at a representation level, the model enhances predictive accuracy and interpretability in tasks such as financial health classification, anomaly detection, and risk signal extraction. Empirical evaluations using publicly available corporate financial reports demonstrate that the proposed approach outperforms single-modality baselines, offering a robust and scalable solution for automated financial statement analysis. The results underscore the potential of combining sequential modeling with contextual language understanding to advance decision-making in financial analytics.