Article: Sentiment analysis and risk early-warning system for cross-border M&A based on natural language processing Journal: International Journal of Computational Vision and Robotics (IJCVR) 2025 Vol.15 No.8 pp.1 - 22 Abstract: In today's volatile global economy, early detection of financial risk is vital for firms engaged in cross-border mergers and acquisitions (M&A). Traditional risk assessment models often struggle to capture the combined effects of financial indicators and qualitative disclosures. This study introduces an intelligent deep learning-based framework, the transfer learning convolutional neural network for financial risk early warning system (TL-CNN-FREWS), to predict financial distress in cross-border M&A firms. The model integrates structured financial metrics with unstructured textual sentiment features to enhance forecasting accuracy. Leveraging transfer learning, a pre-trained VGGNet extracts abstract representations from both data types, while a fine-tuned sequential CNN enables improved risk classification. A robust feature selection pipeline - T-test, RFE-SVM, and random forest - optimises input variables. TLCNN- FREWS effectively captures numerical and linguistic signals, offering timely and accurate financial risk detection. This approach supports decision makers in proactively mitigating risk in complex international financial scenarios. Inderscience Publishers - linking academia, business and industry through research

Title: Sentiment analysis and risk early-warning system for cross-border M&A based on natural language processing

Authors: Yating Chen; Guanying Wei; Eunmi Tatum Lee

Addresses: College of Business and Economics, Sejong University, 209 Neungdong-ro, Gwangjin-gu, Seoul, 05006, South Korea ' College of Business and Economics, Sejong University, 209 Neungdong-ro, Gwangjin-gu, Seoul, 05006, South Korea ' College of Business and Economics, Sejong University, 209 Neungdong-ro, Gwangjin-gu, Seoul, 05006, South Korea

Abstract: In today's volatile global economy, early detection of financial risk is vital for firms engaged in cross-border mergers and acquisitions (M&A). Traditional risk assessment models often struggle to capture the combined effects of financial indicators and qualitative disclosures. This study introduces an intelligent deep learning-based framework, the transfer learning convolutional neural network for financial risk early warning system (TL-CNN-FREWS), to predict financial distress in cross-border M&A firms. The model integrates structured financial metrics with unstructured textual sentiment features to enhance forecasting accuracy. Leveraging transfer learning, a pre-trained VGGNet extracts abstract representations from both data types, while a fine-tuned sequential CNN enables improved risk classification. A robust feature selection pipeline - T-test, RFE-SVM, and random forest - optimises input variables. TLCNN- FREWS effectively captures numerical and linguistic signals, offering timely and accurate financial risk detection. This approach supports decision makers in proactively mitigating risk in complex international financial scenarios.

Keywords: financial distress prediction; transfer learning; sentiment analysis; deep learning; convolutional neural network; CNN; early warning system.

DOI: 10.1504/IJCVR.2025.150869

International Journal of Computational Vision and Robotics, 2025 Vol.15 No.8, pp.1 - 22

Received: 20 May 2025
Accepted: 23 Jul 2025
Published online: 24 Dec 2025 *

Title: Sentiment analysis and risk early-warning system for cross-border M&A based on natural language processing

Keep up-to-date