Browsing by Author "Yogapriya, J"

Now showing 1 - 1 of 1
  • Thumbnail Image
    Item
    AHNet: Design and execution of adaptive hybrid network for credit risk prediction using spatio-temporal attention-based convolutional autoencoder features in the banking sector
    (Springer Nature, 2026-01-07) Ahmad, Ahmad Y. A. Bani; Shukla, Madhu; Jayaprakash, B; Bharathi, B; Ali, Guma; Yogapriya, J
    Over the past ten years, pattern recognition experts have become very interested in market financial predictions. To support investment decision-making processes, it is crucial to develop an intelligent financial forecasting model in the financial markets. However, multivariate financial time series prediction is still difficult. Time series data are typically used for market analysis, and the high degree of fluctuation in this type of data necessitates the use of highly effective classification tools that are prevalent in the state of the art, such as Convolutional Neural Networks (CNNs) systems like AlexNet, residual network, Inception, and so forth. Researchers must start from scratch when training new tools as a result. These procedures could take a long time. Therefore, it is crucial to address the many issues related to using conventional methods for financial forecasting. The benchmark resources are used to gather the necessary financial data for verification in the first step. Next, the collected data are offered to the Spatio-Temporal Attention-based Convolutional Autoencoder (STA-CAE)-based feature extraction phase. The financial prediction phase receives the essential features used for validation extracted in this phase. Here, a novel Adaptive Hybrid Network (AHNet) is employed to perform the prediction procedures. The developed AHNet is an integrated version of Ridge Regression with Stacked Residual Recurrent Neural Network (RR-SResRNN). Moreover, the parameters of AHNet are optimized utilizing the Improved Random Array-based Secretary Bird Optimization Algorithm (IRA-SBOA) that helps to improve the prediction efficiency of the developed prediction technique. The efficiency of the created technique is compared to classical frameworks by executing various tests after collecting financial prediction outcomes from the AHNet model.