MACF-Net: A Multi-Agent Collaborative Financial Intelligence Network for Multimodal SaaS Revenue Forecasting and Risk Assessment

Authors

  • Xingyu Liu University of Chicago, Chicago,IL, USA
  • Peifan Zeng New York University, New York, NY, USA

DOI:

https://doi.org/10.66238/fsrma144

Keywords:

Multi-Agent System, SaaS Financial Analysis, Revenue Forecasting, Risk Assessment, Multimodal Learning, Large Language Model, Financial Intelligence

Abstract

Software-as-a-Service (SaaS) enterprises have increasingly adopted subscription-based business models, resulting in complex financial dynamics characterized by recurring revenue fluctuations, customer lifecycle changes, market uncertainty, and liquidity risks. Accurate revenue forecasting and comprehensive financial risk assessment are therefore essential for SaaS business management and strategic decision-making. However, existing financial intelligence approaches mainly rely on single-model prediction paradigms, which struggle to integrate heterogeneous financial information and provide collaborative reasoning among different analytical perspectives. To address these challenges, this study proposes MACF-Net (Multi-Agent Collaborative Financial Intelligence Network), a multimodal multi-agent framework for SaaS financial analysis, revenue forecasting, and risk assessment. The proposed framework integrates multiple specialized agents, including Revenue Forecasting Agent, Customer Intelligence Agent, Financial Risk Agent, Market Intelligence Agent, and Accounting Verification Agent, which collaboratively analyze financial time-series data, customer behavior information, enterprise documents, and external market signals. Furthermore, a Cross-Agent Evidence Verification Mechanism and a Conflict-aware Agent Debate Mechanism are developed to improve decision reliability by enabling evidence sharing and dynamic prediction refinement among agents. Experimental results on the Multimodal SaaS Financial Intelligence Dataset (MSFID) demonstrate that MACF-Net achieves superior performance compared with existing baselines. Specifically, MACF-Net reduces the Mean Absolute Percentage Error (MAPE) of revenue forecasting from 7.31% to 6.02% and improves the Area Under the Receiver Operating Characteristic Curve (AUC) for financial risk assessment from 0.914 to 0.962. These results verify the effectiveness of collaborative multi-agent reasoning for SaaS financial intelligence.

References

[1] Zhang, Bingli, et al. "Consensus Matrix: A Role-Specialized Multi-Agent Framework for Structured Collaborative Decision-Making in Agentic Visual Media Workflows." Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 2026.

[2] Chen, Wen. "Enterprise financial risk prediction and intelligent early warning model based on deep learning." Discover Artificial Intelligence 5.1 (2025): 227.

[3] Navon, Ariel, and Yosi Keller. "Financial time series prediction using deep learning." arXiv preprint arXiv:1711.04174 (2017).

[4] Lim, Bryan, et al. "Temporal fusion transformers for interpretable multi-horizon time series forecasting." International journal of forecasting 37.4 (2021): 1748-1764.

[5] Wu, Haixu, et al. "Autoformer: Decomposition transformers with auto-correlation for long-term series forecasting." Advances in neural information processing systems 34 (2021): 22419-22430.

[6] Ding, Xiao, et al. "Deep learning for event-driven stock prediction." Ijcai. Vol. 15. 2015.

[7] Hossain, Md Refat, et al. "Enhancing Credit Scoring with Multimodal Deep Learning: A Hybrid Neural Network Approach Using Structured and Unstructured Financial Data." International Interdisciplinary Business Economics Advancement Journal 6.07 (2025): 09-22.

[8] Araci, Dogu. "Finbert: Financial sentiment analysis with pre-trained language models." arXiv preprint arXiv:1908.10063 (2019).

[9] Wu, Shijie, et al. "Bloomberggpt: A large language model for finance." arXiv preprint arXiv:2303.17564 (2023).

[10] Yao, Shunyu, et al. "React: Synergizing reasoning and acting in language models." arXiv preprint arXiv:2210.03629 (2022).

[11] Schick, Timo, et al. "Toolformer: Language models can teach themselves to use tools." Advances in neural information processing systems 36 (2023): 68539-68551.

[12] Ding, Han, et al. "Large language model agent in financial trading: A survey." Proceedings of the 2026 9th International Conference on Computers in Management and Business. 2026.

[13] Park, Joon Sung, et al. "Generative agents: Interactive simulacra of human behavior." Proceedings of the 36th annual acm symposium on user interface software and technology. 2023.

[14] Guo, Taicheng, et al. "Large language model based multi-agents: A survey of progress and challenges." arXiv preprint arXiv:2402.01680 (2024).

[15] Chen, Shuaihang, et al. "A survey on llm-based multi-agent system: Recent advances and new frontiers in application." arXiv preprint arXiv:2412.17481 (2024).

[16] Hua, Zonghui, Yue Luo, and Yunting Ling. "LPG-SAC-TCO: A Large Language Model--Policy-Guided Soft Actor-Critic Framework for Total Cost of Ownership--Optimal Energy Management of Hydrogen Fuel Cell Heavy-Duty Trucks." European Journal of AI, Computing & Informatics 2.3 (2026): 1-11.

[17] Ling, Yunting, and Wenxuan Liu. "SR-DTMA: A Digital Twin-Driven LLM Multi-Agent Framework for Systemic Risk Simulation and Coordinated Decision-Making in Supply Chains." Journal of Computer, Signal, and System Research 3.1 (2026): 123-132.

[18] Li, Jiawei, Yue Luo, and Yunting Ling. "Carbon-Aware Reinforcement Learning for Cost-Optimal Energy Management of Hydrogen-Electric Heavy-Duty Trucks." Strategic Management Insights 3.2 (2026): 1-10.

[19] Zhang, Jingsi, and Jixiang Ding. "CA-PPO-AXI: Coverage-Aware Deep Reinforcement Learning for Adaptive AXI Bus Protocol Verification." Fundamental Scientific Reports in Multidisciplinary Areas 2.02 (2026): 159-169.

[20] Zhao, Qinhe, and Yuheng Huang. "DAREA: A Dynamic AI-Driven Architecture for Real Estate Asset Tokenization on Blockchain." Fundamental Scientific Reports in Multidisciplinary Areas 2.02 (2026): 182-192.

[21] Li, Siyu, Rongshuang Sun, and Jinlin Hu. "Adaptive risk control model: A customer credit default detection framework based on random forest and neural network." Journal of Technology Innovation and Engineering 1.3 (2025).

[22] Qiu, Min, Linting Wu, and Shanshan Wang. "CyberDQFormer: A Knowledge Graph-Enhanced Transformer Framework for Security-Aware Data Quality Anomaly Detection and Autonomous Repair Against Network Attacks." Fundamental Scientific Reports in Multidisciplinary Areas 2.3 (2026): 34-52.

Downloads

Published

2026-10-31