Jul 2026· International Conference Computing Methodologies and Communication· pp. 1604-1609· 0 citations· 20 references
Abstract
Standard credit-risk scorecards rely on linear ratio thresholds that break down when feature interactions are nonlinear and observations carry temporal dependencies. Qualitative signals embedded in corporate disclosures—tone shifts, forward-looking hedges, and sector-specific terminology—remain largely ignored by numeric-only models, even though such signals often precede ratio deterioration. This paper introduces a tri-modal deep learning framework that jointly trains three complementary branches: a Convolutional Neural Network (CNN) for cross-sectional ratio-pattern detection, a Long Short-Term Memory (LSTM) network for multi-quarter trend modelling, and a Natural Language Processing (NLP) branch for disclosure-text encoding. Prior to deep-model training, LASSO regularisation removes collinear financial indicators and SMOTE oversampling corrects the severe class imbalance characteristic of distress datasets. A feature-concatenation fusion layer integrates all three branch outputs; the resulting vector feeds a sigmoid classifier that produces a calibrated distress probability. Benchmarked against five baselines on four financial datasets, the model reaches 94.8% accuracy and 91.3% minority-class recall, with a 4.1-point F1 advantage over the strongest single-modality competitor.
An Integrated Data-Driven Loan Management Framework that comprises an Artificial Neural Network (ANN) for credit default prediction, Structured Query Language (SQL) to systematically extract and transform data, and Interactive Visual Analytics Dashboards to aid in providing transparency within the decision-making process is discussed.
Ponsak .S. Bande, Chimuzuoroke E. Ugwuja, Blessing .C. Uzo et al.· International Journal of Lat...· 0 citations
Financial risk assessment for manufacturing enterprises supports credit-risk screening and early warning. Existing methods often fuse multi-period financial and industry information through concatenation, shared representations, or unrestricted interactions, making it difficult to separate meaningful disturbance–sensitivity correspondences from irrelevant combinations. This study proposes an industry-disturbance-conditioned credit-scoring framework. Its originality lies in explicitly matching external disturbances with firm-level financial sensitivities rather than treating them as unrestricted features. The framework decomposes financial information into levels, intertemporal changes, and accounting divergences; constructs demand, cost, and production sensitivities; and estimates conditioned exposures through a correspondence matrix and conditional gates. A matching-consistency loss constrains disturbance–sensitivity relationships, while dual-path prediction retains exposure-related and firm-specific risk information. Using Moody’s Orbis and Eurostat Short-Term Business Statistics (STS), the method achieves an AUPRC of 0.512 in the full out-of-time test, exceeding TabPFN, the strongest AUPRC baseline, by 1.4 percentage points. Its AUROC is 0.879, and its FNR of 0.276 is lower than 0.289 for HGNN and 0.291 for TabPFN. Among highly sensitive firms, the proposed model achieves an AUPRC of 0.489 and an FNR of 0.288, improving on HGNN by 1.5 and 2.3 percentage points, respectively. Statistical tests, ablation studies, and repeated runs support the matching and prediction mechanisms. Independent calibration further reduces probability error and calibration bias. The framework supports relative risk ranking and distress screening during industry disturbances, although validation across additional databases and disturbance settings remains necessary.
Shi-Liang Chang· ICST Transactions on Scalabl...· 0 citations
Accurately predicting stock returns remains one of the core challenges in quantitative finance, exacerbated by nonlinear market dynamics, irrational investor behavior, and the stochasticity of financial time series. This paper proposes a novel hybrid framework—Bayesian Optimized Long Short-Term Memory Network Based on Sentiment Enhancement Technical Indicators (BOLSTM-SATI), which combines a Long Short-Term Memory (LSTM) neural network with classic technical analysis signals and multi-source investor sentiment indices from the CSI 300 index constituents. This study constructs a composite feature space including MACD, RSI, Bollinger Bands, volume balance indicator, and Average True Range (ATR). Investor sentiment is quantified through a Bayesian update mechanism that integrates news text sentiment scores obtained through FinBERT and social media sentiment indices from the Eastmoney Guba forum. To mitigate overfitting and improve generalization ability, we employ a Bayesian optimization method based on a tree-structured Parzen estimator (TPE) for LSTM hyperparameter search. Robustness to perturbations is verified through Gaussian noise injection experiments. Empirical evaluation using daily data from January 2015 to December 2022 shows that BOLSTM-SATI achieves a directional accuracy of 67.3%, a mean absolute error (MAE) of 0.0082, and a Sharpe ratio of 2.41 in backtesting simulations, outperforming baseline models including ARIMA, traditional LSTM, XGBoost, and Transformer variants. The results confirm that incorporating sentiment dynamics into technical feature learning can significantly improve prediction accuracy and the profitability of trading strategies.
Bairui Peng· Advances in Economics, Manag...· 0 citations
Predicting financial asset returns remains one of the most difficult challenges in empirical finance, driven by the low signal-to-noise ratio and the semi-strong form of market efficiency. While deep learning models, especially LSTM networks, have shown promise in capturing temporal dependencies, standard architectures often struggle to account for the cross-sectional heterogeneity of asset returns. This paper proposes a novel architectural extension to the basic LSTM model designed to improve both predictive accuracy and model interpretability. The framework integrates macro-financial covariates to capture broader economic signals and learnable sector embeddings to encompass heterogeneity by sector. The trading strategy involves constructing a long-short portfolio based on daily directional forecasts for each S&P 500 constituent, targeting stocks expected to under- or outperform the cross-sectional median return of the S&P 500. Model Performance is evaluated against three competitive benchmarks: a basic LSTM, a Random Forest model and a traditional market buy-and-hold strategy. The empirical results demonstrate that the LSTM with sector embeddings outperforms all benchmarks across key risk and return metrics. By utilizing sector embeddings, the model explicitly incorporates cross-sectional heterogeneity, allowing it to adapt to varying industry dynamics within the market. To address the black-box nature of deep learning, I use latent space visualizations to analyse how the model differentiates between sectors, providing insights into the internal representation of the sectors in the LSTM. The impact of the sector information can be quantified using a novel contribution metric by inspecting the weights of the LSTM. The predictive signal is driven by a short-term reversal factor and an industry momentum factor.
Short-horizon AUD/USD direction forecasting is a noisy classification problem relevant to risk monitoring and foreign-exchange exposure management. Many exchange-rate forecasting studies evaluate model classes after combining technical, financial, and macro-related predictors, which can obscure whether performance differences arise from the learner itself or from economic information channels. This paper uses a paired feature-group diagnostic design that combines a full-feature benchmark, OwnMarket-based inclusion, AllGroups-based ablation, window-length sensitivity, and expanding-window robustness analysis. The 45 daily predictors are grouped as OwnMarket, CrossFX, Commodities, EquityRisk, and DollarRatesSpreads. After conservative lag-1 alignment, the best full-feature MCC is 0.076, achieved by the 1D-CNN. Validation-based inclusion-ablation diagnostics rank EquityRisk first among the external groups, driven mainly by its positive ablation contribution. Commodity effects are mixed across models, lookback windows, and expanding-window periods, while overall directional skill remains limited. The results show that paired feature-group diagnostics provide a disciplined way to interpret information-source contributions when model-only rankings are weakly informative.
Hengren Xu, Jinchuan Zheng, Zhenwei Cao et al.· International Conference on...· 0 citations
Accurate credit risk prediction plays a critical role in strengthening the financial stability of microfinance institutions, especially in developing economies, where increasing loan defaults and imbalanced borrower records create significant challenges for reliable decision-making. Although machine learning approaches have improved credit assessment practices, existing models often favor majority-class borrowers and fail to detect high-risk default cases effectively because of severe class imbalance. This limitation highlights the need for more robust and imbalanced-sensitive predictive frameworks. This study aims to develop an effective machine learning-based credit risk prediction framework by integrating data balancing strategies with ensemble and deep learning models. This study systematically investigates the impact of baseline learning and multiple resampling techniques, including oversampling, undersampling, and hybrid methods, when applied to Random Forest, XGBoost, LightGBM, CatBoost, and Deep Neural Network classifiers. The effectiveness of the proposed models was assessed using imbalance-aware evaluation measures, particularly ROC-AUC and Geometric Mean, along with conventional classification metrics. The experimental findings demonstrate that incorporating resampling techniques substantially improved the default detection performance. The DNN model combined with SMOTETomek achieved the best results, obtaining 94.9% F1-score, ROC-AUC 98%, and 97.2% of G-Mean. CatBoost also exhibited consistent competitiveness across different sampling configurations. These findings suggest that hybrid sampling integrated with advanced learning architectures can provide a reliable and practical solution for managing credit risk in imbalanced microfinance datasets, supporting improved lending decisions and sustainable financial operations in the future.
