面向互聯網輿情事件的企業風險識別

張志劍 劉政昊 馬費成

摘 要：信息智能時代背景下，互聯網輿情信息對企業的影響愈加顯著。有效準確地從輿情事件中識別風險有助于企業進行風險管理，實現良性運營。本文提出一種有效識別企業風險的模型KGANN，該模型使用知識圖譜的結構和內容構造神經網絡，實現知識圖譜和神經網絡的融合，從而提升模型風險識別能力。實驗結果表明，在企業風險識別任務上所提方法相較于傳統方法具有顯著優勢。同時根據知識的權重值對模型進行分析，得到股權結構復雜、司法案件較多、知識產權較少的企業風險等級較高。研究結果為企業和監管機構進行風險管理提供了重要的研究視角，對防范企業風險具有一定的參考價值。

關鍵詞：互聯網輿情;風險識別;風險事件;知識圖譜;神經網絡

中圖分類號：F272.35 文獻標識碼：A 文章編號：2097-0145（2022）01-0065-09 doi：10.11847/fj.41.1.65

Abstract：Under the background of the information intelligence era， the impact of Internet public opinion information on enterprises is becoming more and more significant. Effectively and accurately identifying risks from public opinion events is helpful for enterprises to carry out risk management and realize benign operations. This paper proposes a model KGANN for effectively identifying enterprise risk. The model uses the structure and content of knowledge graph to construct a neural network to realize the integration of knowledge graph and neural network to improve the ability of model risk identification. The experimental results show that the proposed method has significant advantages over the traditional methods in enterprise risk identification. At the same time， the model is analyzed according to the weight value of knowledge. It is concluded that the enterprise with a complex ownership structure， more judicial cases， and less intellectual property rights has a higher risk level. The research results provide an essential research perspective for enterprises and regulators to carry out risk management and have a specific reference value for preventing enterprise risks.

Key words：internet public opinion; risk identification; risk events; knowledge graph; neural network

1 引言

近年來，隨著經濟全球化進程的不斷加速，市場主體各要素間的關聯性不斷增強，金融系統的風險敞口也日益增大。在全球經濟貿易互通互聯、國內經濟加速轉型的特殊階段，企業正處于創新發展與經濟轉型升級的關鍵時期，然而由于企業內外環境的不確定性、生產經營活動的高度復雜性和部分企業能力的有限性，導致各類風險因子高度集中，企業風險事件頻發[1]。與此同時，互聯網媒體的發展大大加快了各行業信息產生和傳播速度，一些負面的互聯網輿情事件的爆發與傳播更是加劇了企業風險，給企業帶來了巨大的財產損失和聲譽損失。……