Tourism demand forecasting has evolved into a wide variety of models, including time-series models that incorporate economic, environmental, and behavioral factors. Macao, one of the world’s most profitable gaming destinations, finds that gaming revenue is highly related to tourist arrivals. A forecast model for gaming tourism is essential for accurately predicting tourist arrivals. The challenge with ARIMA-type models is optimizing parameter selection in order to improve the accuracy of tourism demand forecasts. In this study, an enhanced version of SARIMAX, called SARIMAX-E, was developed to identify the most effective parameter combinations. By integrating data related to gaming revenue, weather, transportation, currency exchange rate, holidays, and seasonality into a single forecast model, this study examined the performance of different forecasting models, including the proposed SARIMAX-E model; ARIMA-type models (ARIMA, SARIMA, ARIMAX); and machine learning models (Transformer, LTSM, Random Forests, XGBoost). The results showed that the ARIMA family of models, including SARIMAX-E, ARIMAX, and SARIMA, was particularly well suited to tourism demand forecasting, as its members consistently ranked among the top performers in terms of error metrics. By applying multi-step predictions, LSTM outperforms most conventional approaches. Compared with all other models, the SARIMAX-E performed the best after applying the additional parameter grid.

Traditional financial early warning models often rely on the independent and identically distributed (IID) assumption, failing to adequately capture cross-sectional spatial contagion effects and temporal dynamic mutations, and are susceptible to the over-smoothing problem when processing highly imbalanced graph networks. To address these limitations, this study proposes a micro-manifold-based identity-preserving spatiotemporal graph neural network framework (Micro-STAGNN). In the spatial dimension, an identity-preserving graph convolutional operator (IP-GCN) is constructed. By hard-coding a self-preservation coefficient ($\mathsf{\lambda }=0.8$), it quantifies peer risk spillover while mitigating feature dilution, ensuring the transmission of heterogeneous default signals. In the temporal dimension, Long Short-Term Memory networks are cascaded with a temporal attention mechanism to capture the nonlinear temporal inflection points that trigger financial distress. The empirical study utilizes a sample of China’s A-share market from 2015 to 2025, evaluating the model using an Out-of-Time Validation protocol and Focal Loss. Results indicate that under a highly imbalanced distribution with a positive-to-negative sample ratio of approximately 1:50, Micro-STAGNN achieves an OOT ROC-AUC of 0.9095, a minority class default recall of 89%, and reduces the missed detection rate to 11%, outperforming traditional nonlinear cross-sectional models such as XGBoost. Furthermore, temporal attention weights provide explainable support for the early warning results.

In classical vernacular Chinese fiction, inferring affective stance in the presence of rhetorical devices often depends on implicit semantics and contextual cues, which poses substantial challenges for joint pragmatic modeling with large language models (LLMs). To address these challenges, we propose KnowRhet, a unified framework for joint pragmatic modeling in classical vernacular Chinese fiction. Under a unified structured-output interface, KnowRhet integrates task-specific domain knowledge guidance, reasoning-process control, and LoRA-based parameter-efficient adaptation, thereby improving the ability of LLMs to jointly identify rhetorical devices and affective stance. Here, task-specific domain knowledge refers to domain-specific interpretive knowledge operationalized through label definitions, category descriptions, and boundary constraints for the current literary analysis setting, rather than knowledge obtained from external retrieval. To evaluate the proposed framework, we construct a sentence-level benchmark based on Dream of the Red Chamber, comprising 2,769 manually annotated instances for multi-label rhetorical device prediction, emotion expression type classification, and sentiment polarity classification. Experimental results show that explicit label definitions, category descriptions, and boundary constraints yield the most consistent gains, improving Avg F1 from 58.53% to 75.31%. Structured reasoning exhibits clear task-dependent effects, while LoRA improves Qwen2.5-7B-Instruct from 57.05% to 75.39% Avg F1, approaching the prompt-only performance of DeepSeek-V3.2 (75.77%) under the same evaluation protocol. These findings suggest that robust pragmatic modeling in classical vernacular Chinese fiction depends less on increasing reasoning complexity alone than on combining explicit task constraints with lightweight domain adaptation.

Accurate multivariate long-term time series forecasting (LTSF) is critical for smart grid operations. However, non-stationary distribution shifts frequently induce compounding error accumulation in conventional architectures. This study proposes the Mamba-LSTM-Attention (MLA) framework, a distribution-aware architecture engineered for forecasting stability. The pipeline integrates Reversible Instance Normalization (RevIN) to neutralize statistical drift. To address computational bottlenecks, the architecture utilizes a linear-time Selective State Space Model (Mamba) to capture global trend dynamics, cascaded with a single-layer gated Long Short-Term Memory (LSTM) unit to model localized non-linear residuals. A terminal information bottleneck structurally bounds cross-step error propagation. Empirical results across standard ETT and Electricity benchmarks reveal a precision–stability trade-off. By prioritizing structural resilience, the MLA framework limits error accumulation on highly volatile datasets, yielding MSEs of 0.210 and 0.128 on ETTh2 and ETTm2 at the T = 96 horizon. This structural bottleneck inherently smooths high-frequency periodic patterns, yielding lower absolute accuracy on stationary benchmarks such as ETTh1 and ETTm1. Ultimately, the architecture establishes a computationally efficient, structurally stable baseline tailored for non-stationary anomaly tracking in smart grids.

Traditional financial early warning models often rely on the independent and identically distributed (IID) assumption, failing to adequately capture cross-sectional spatial contagion effects and temporal dynamic mutations, and are susceptible to the over-smoothing problem when processing highly imbalanced graph networks. To address these limitations, this study proposes a micro-manifold-based identity-preserving spatiotemporal graph neural network framework (Micro-STAGNN). In the spatial dimension, an identity-preserving graph convolutional operator (IP-GCN) is constructed. By hard-coding a self-preservation coefficient (λ=0.8), it quantifies peer risk spillover while mitigating feature dilution, ensuring the transmission of heterogeneous default signals. In the temporal dimension, Long Short-Term Memory networks are cascaded with a temporal attention mechanism to capture the nonlinear temporal inflection points that trigger financial distress. The empirical study utilizes a sample of China's A-share market from 2015 to 2025, evaluating the model using an Out-of-Time Validation protocol and Focal Loss. Results indicate that under a highly imbalanced distribution with a positive-to-negative sample ratio of approximately 1:50, Micro-STAGNN achieves an OOT ROC-AUC of 0.9095, a minority class default recall of 89%, and reduces the missed detection rate to 11%, outperforming traditional nonlinear cross-sectional models such as XGBoost. Furthermore, temporal attention weights provide explainable support for the early warning results.

Forecasting long-term time series (LTSF) requires a delicate balance between capturing global dependencies and preserving granular local dynamics. Although State Space Models (SSMs) and Transformers have been successful, a technical challenge remains, in which individual paradigms often fail to perform well over extended horizons due to the difficulty of simultaneously achieving linear-time efficiency and high-fidelity local refinement at the same time. This study introduces Mamba-LSTM-Attention (MLA), a novel hybrid architecture featuring a cascaded topology designed to bridge these dimensions. In this work, the core innovation is the hierarchical feature evolution mechanism, in which a Mamba module serves first as an efficient global encoder to capture long-range periodic trends at linear complexity. Following this, gated LSTM units are used to refine the algorithm at the micro-scale in order to filter noise and characterize non-linear local fluctuations. Lastly, a multi-head attention mechanism performs a dynamic feature re-weighting in order to focus on key historical signals. Systematic evaluations across four multivariate benchmark datasets demonstrate that MLA achieves exceptional cross-step forecasting stability. Most notably, on the ETTh1 dataset, MLA maintains a remarkably narrow Mean Squared Error (MSE) fluctuation range (0.127) as the forecasting horizon extends from T=96 to T=720. This empirical evidence confirms that the integrated Mamba module effectively mitigates the error accumulation typically encountered by vanilla LSTMs. While the current implementation faces an information bottleneck due to a single-point projection decoding strategy, the ablation studies (revealing a 19.76% MSE surge upon LSTM removal) validate the combination of the proposed architecture. This work establishes a robust framework for hybrid SSM-RNN modeling and a clear path for future performance enhancements through sequence-to-sequence mechanisms.