ABSTRACT The identification of personal credit risk constitutes a fundamental concern within the realm of financial risk management. As the credit industry experiences significant growth, the precise evaluation of borrowers' credit risk and the mitigation of credit default risk have emerged as critical priorities for financial institutions and researchers worldwide. To enhance the ability to identify defaulting customers, this paper proposes a credit risk identification algorithm based on Bagging Fast Correlation‐Based Filter with Temporal Convolutional Network (BaggingFCBF‐TCN). This algorithm initially incorporates the feature selection approach inherent in the Bagging strategy to identify and filter the characteristics associated with defaulting customers, which serves to mitigate the bias in feature selection outcomes that may favor the majority class. Subsequently, it employs an enhanced Temporal Convolutional Network (TCN) classifier for the purpose of credit risk assessment, thereby improving the ability to discern both long‐term and short‐term dependencies present in personal credit data. The test results show that: (1) The BaggingFCBF‐TCN algorithm significantly enhances the model's ability to identify defaulting customers, achieving optimal overall identification performance. (2) The results of the combination effect analysis indicate that the personal credit risk identification model constructed using the BaggingFCBF‐TCN combination algorithm outperforms other combination algorithms in both the original dataset and the dataset after class balancing treatment.

ABSTRACT Carbon emissions are a significant contributor to global warming. As one of the largest carbon emitters in the world, China is committed to establishing a carbon emission trading market to address the challenges posed by climate change. The carbon price is a fundamental component of the carbon financial market. Accurately predicting it can improve environmental quality, reduce energy demand, and promote economic growth. This study uses price data from the Guangdong carbon market as a case study and employs a hybrid model that integrates Convolutional Neural Networks (CNN) and Long Short‐Term Memory (LSTM) networks for carbon price forecasting. The findings indicate that: (1) the CNN–LSTM model exhibits optimal predictive performance when the sliding window is set to a size of 5 on the basis of previous carbon price data. (2) By incorporating significant indicator features from the Guangdong pilot carbon price dataset while maintaining a sliding window size of 5, the model achieves superior predictive accuracy, as evidenced by a Goodness of Fit (R2) of 0.8622 and a mean absolute error (MAE) of 0.0228, resulting in the most favorable comprehensive evaluation index. (3) The integration of one‐dimensional convolutional layers with LSTM layers in the CNN–LSTM model effectively leverages the strengths of CNNs for local feature extraction and the capabilities of LSTMs for modeling time series data. This approach leads to a substantial improvement in predictive performance compared with alternative models such as Support Vector Machine (SVM), Recurrent Neural Network (RNN), and LSTM.

The widespread global adoption of credit card payments has significantly increased convenience for consumers. However, this shift has also heightened the demand for robust fraud detection systems across various financial institutions. While numerous fraud detection algorithms have been proposed in both academia and industry, significant challenges remain in the accurate identification of minority class samples and the effective capture of time series information during feature extraction. To address these challenges, this paper proposes a long short-term memory-logistic regression stacked ensemble model (LLSE) that integrates long short-term memory (LSTM) and logistic regression (LR). The first layer of the model extracts time series features by combining three independently trained LSTM base learners whose outputs are concatenated and passed to the second layer, an LR meta-learner, to generate the final predictions. This dual-layer decision-making mechanism effectively captures spatiotemporal correlation features in fraudulent transactions, thereby enhancing the overall performance of the model. The model’s performance is evaluated via two publicly available credit card datasets. This study innovatively combines the TS-N2VA feature extraction model with the LLSE classifier to comprehensively enhance credit card fraud detection performance through both hidden feature extraction and classifier improvement. Experiments on two public credit card datasets demonstrate that the LLSE classifier achieves significantly better performance in identifying minority class samples while maintaining overall recognition accuracy, outperforming other fraud detection models in the experiments.

Rolling bearings fault diagnosis technology plays a significant role in modern industries. In this article, a novel method to diagnose bearing faults via acoustic metamaterials is proposed and experimentally validated by selecting and enhancing the optimal resonance frequency band for demodulation. The proposed method is performed by analysing acoustic signals, which is convenient due to its noncontact measurement. Different from conventional denoising techniques, the proposed method is implemented by enhancing the resonance frequency band via acoustic metamaterials instead of constructing filters to wipe out noise. Consequently, all the useful information of fault signals are remained and the detection limits of current acoustic sensing systems are improved because of the enhanced fault signals. Additionally, the optimal resonance band is sought by changing measured positions based on the frequency-selective enhancement property of the designed metamaterial. By comparing with spectral kurtosis-based method, the proposed method is more superior to extract bearing fault features at low signal to noise ratios. All the presented results indicate that the proposed method is effective to extract bearing fault features.

With the development of information technology, the Internet has become an important channel of public opinion for expressing public interests, emotion, and ideas. Public emergency usually spreads via network. Due to the temporal and spatial flexibility and the information amplification of network, the opinions from different regions and background are easy to be represented as network public opinion, and have important impact on social and economic life. Thus, studying the formation mechanism of network public opinion has important theoretical and practical significance. Taking the formation process of network public opinion under emergencies as the research object, this paper first identifies the key factors influencing the formation of network public opinion, namely the internal characteristics (include individual education level, individual stubbornness, individual initial opinion, and so on) and external information of individuals (include external information intensity). Second, information intensity is introduced to describe the influence of external information feature on the formation of network public opinion. Individual education level, individual stubbornness, and individual initial opinion are analyzed to describe the influence of individual internal factors on the formation, and then its model is constructed. Through the simulation experiments, this paper analyzes the influence of external information intensity, individual education level, individual stubbornness, individual initial opinion, and other factors on the formation of network public opinion. The simulation results show that: (1) the greater intensity of public emergency reporting causes the easier formation of network public opinion; (2) the higher individual education level leads to the shorter time for completing the final formation and stable state of online public opinions, and after the formation of online public opinions, the opinion of the event is mainly neutral; (3) the greater individual's stubbornness makes the shorter formation time of online public opinion. When online public opinion reaches a stable state, the neutral opinion group dominates and firmly controls the development trend of public opinion; (4) the difference of opinions among individuals is the most important factor affecting the formation of network public opinion. Finally, the rationality and validity of the proposed model are verified by a real case. Compared with previous studies on the formation mechanism of network public opinion, this paper divides the formation process of network public opinion into three stages: individual information perception, individual decision making, and individual opinion transmission. Meanwhile, the influence of individual internal factors and external information characteristics on the formation process of network public opinion is also considered.

Summary The ant colony labor division model can be used to solve the dynamic and variable traffic signal timing problem because of its adaptive and self‐adjusting characteristics. Based on the basic model, this paper proposes a new extended ant colony labor division model for traffic signal timing. This model is combined with the vehicle characteristics to modify the calculation method of environmental stimulus values. Using the vehicle delay in the unit period, the original fixed response threshold is modified to the dynamic response threshold, and the state transition equation is reconstructed. Based on the mixed traffic flow model of two‐phase single‐point intersection cellular automata, through comparative experiments and discussion and analysis, it is found that the extended ant colony labor division model can effectively improve road traffic capacity according to road conditions and reasonable traffic signal timing.

Summary Currently, group behaviors happen frequently with the development of network technology. As a typical social group behavior, group polarization has been attracted more and more academic attention due to its significant disturbance to public's daily lives. At present, the classic J‐A (proposed by Jager and Amblard) and D‐W (proposed by Deffuant and Weisbuch) models are used to analyze group polarization process. However, the main shortcomings existing in these models are that the individuals' psychology and their network relationships are rarely considered. In order to overcome the limitations, this article integrates the influence factors such as conformity and network relationship strength integrated into the polarization model. Besides, the BA (proposed by Barabasi and Albert) network model is used as the agent adjacency model due to its closer to the real social network structure. Subsequently, the experimental simulations are carried out with the multi‐agent Monte‐Carlo method so as to testify the efficiency and effectiveness. The results indicate that different information interaction modes have essential influence on group attitude polarization. Moreover, conformity parameters and the intensity of relationship have dual impacts on both speeding up and slowing down the polarization process.