In the era of Artificial Intelligence and Big Data, financial fraud is inevitable. This article proposes ENKMRH, a novel selective ENsemble prediction method based on K-Means++ and the Refractive Inverse Learning Harris Hawks Optimization algorithm (RILHHO), for combatting fraud risk in digital financial systems, designed to adapt to complex, high-dimensional, and nonlinear financial data. First, we innovatively apply K-means++ for the diversified selection of well-performing base learners and employ distance-based selection to preliminarily select partial learners with better overall performance. This approach conserves computing resources and improves the generalization ability and stability of the ensemble system. Second, we introduce chaos initialization, an improved position update, an escape energy strategy based on biological principles, and RILHHO. RILHHO is designed to provide an efficient and precise selection strategy for the selective ensemble prediction of digital financial fraud risk. Finally, we apply the proposed model to three typical real-world problems in digital financial fraud: internet consumer credit fraud, online lending fraud, and money laundering risk prediction. The experimental results demonstrate that ENKMRH outperforms other state-of-the-art basic techniques and ensemble learning models, achieving the highest accuracy rates of 81.39%, 88.68%, and 93.80% across three financial fraud datasets. The research findings offer crucial guidance to financial practitioners, aiding in investment decisions and bolstering financial stability and security. Furthermore, they enhance institutions’ risk management capabilities, fostering sustainable growth and prosperity. This article intertwines financial risk management with the domains of technology and engineering management, utilizing advanced algorithms and data analytics techniques to tackle modern challenges in digital financial systems.
11 May 2022 in Arabian Journal for Science and Engineering
The frequent bankruptcy incidents of peer-to-peer (P2P) lending industry have damaged the benefits of investors in China, and how to accurately and efficiently predict the investment risks of P2P lending becomes an important problem. For this very reason, a novel prediction method based on improved binary glowworm swarm optimization and multi-fractal dimension (IBGSOMFD) for P2P lending investment risk is proposed. Firstly, we propose an improved binary glowworm swarm optimization, abbreviated IBGSO, by uniformly designing an initial population using the good-point set theory, improving the moving way of glowworms, and introducing the mechanism of population diffusion and variation. Secondly, IBGSO combined with multi-fractal dimension (MFD) is applied to feature selection, and the optimal subset extracted from the original dataset can be efficiently achieved utilizing IBGSO, which can reduce its redundant attributes, and retain its pivotal attributes of P2P lending investment risk. Finally, an investment risk prediction model of P2P lending based on support vector machine (SVM) is established, which can accurately and efficiently predict the investment risk of P2P lending. Experimental results on 6 University of California Irvine (UCI) benchmark datasets show that IBGSOMFD outperforms other state-of-the-art approaches in predictive ability and calculative efficiency, and its effectiveness and significance. After the performance verification of IBGSOMFD, this work looks at how it can be applied in the risk prediction of P2P lending investment in China to maintain a stable market order.
Accurate traffic flow prediction plays a crucial role in the development of intelligent traffic management. Despite numerous investigations into spatio-temporal methods, achieving high accuracy in traffic flow prediction remains challenging. This challenge arises from the complex dynamic spatio-temporal correlations within the traffic road network and the limitations imposed by the selection of hyperparameters based on experiments and manual experience, which can affect the performance of the network architecture. This paper introduces a novel transformer-based adaptive graph convolutional recurrent network. The proposed network automatically infers the interdependencies among different traffic sequences and incorporates the capability to capture global spatio-temporal correlations. This enables the dynamic capture of long-range temporal correlations. Furthermore, the whale optimization algorithm is employed to efficiently design an optimal network structure that aligns with the requirements of the traffic domain and maximizes the utilization of limited computational resources. This design approach significantly enhances the model’s performance and improves the accuracy of traffic flow prediction. The experimental results on four real datasets demonstrate the efficacy of our approach. In PEMS03, it improves MAE by 2.6% and RMSE by 1.4%. In PEMS04, improvements are 1.6% in MAE and 1.4% in RMSE, with a similar MAPE score to the best baseline. For PEMS07, our approach shows a 4.1% improvement in MAE and 2.2% in RMSE. On PEMS08, it surpasses the current best baseline with a 3.4% improvement in MAE and 1.6% in RMSE. These results confirm the good performance of our model in traffic flow prediction across multiple datasets.
Edge computing enables efficient data aggregation for services like data sharing and analysis in distributed IoT applications. However, uploading dynamic high-dimensional data to an edge server for efficient aggregation is challenging. Additionally, there is the significant risk of privacy leakage associated with direct such data uploading. Therefore, we propose an edge-based differential privacy data aggregation method leveraging progressive UMAP with a dynamic time window based on LSTM (EDP-PUDL). Firstly, a model of the dynamic time window based on a long short-term memory (LSTM) network was developed to divide dynamic data. Then, progressive uniform manifold approximation and projection (UMAP) with differential privacy was performed to reduce the dimension of the window data while preserving privacy. The privacy budget was determined by the data volume and the attribute’s Shapley value, adding DP noise. Finally, the privacy analysis and experimental comparisons demonstrated that EDP-PUDL ensures user privacy while achieving superior aggregation efficiency and availability compared to other algorithms used for dynamic high-dimensional data aggregation.