忘却学习视角下企业数字化转型的多重路径分析——基于机器学习算法

The Multiple Paths of Enterprise Digital Transformation from the Perspective of Unlearning:Based on machine learning algorithms

摘要:

在数字经济浪潮下，技术和管理模式等知识的更替使得忘却学习被视为企业实现数字化转型的有效途径，但相关研究缺乏识别不同企业特征及其在数字化转型不同阶段所采取的忘却学习方式。因此，为对企业数字化转型提供具有针对性的理论与实践参考，基于忘却学习的视角，将行业数字化程度、地区数字经济发展水平以及企业年龄作为企业类型划分依据，采用K-means算法将2017－2021年我国开展数字化转型的A股上市企业划分为行业和地区双赢型、行业优势型、地区优势型3类不同群组，将战略制定、转型实施、评估迭代作为企业数字化转型的3个阶段，运用轻量级梯度提升机算法（LightGBM）获取不同类型企业在转型各阶段的学习方式并组合为其数字化转型的学习路径。结果显示：在开展数字化转型时，行业和地区双赢型企业优先注重外部知识的获取并逐渐将重心转移到过时知识的抛弃，行业优势型企业全程注重外部知识的获取，地区优势型企业优先注重内部过时知识的抛弃并逐渐将重心转移到外部知识的获取；此外，数字化转型对行业追随者企业具有很强的吸引力。因此，对于计划数字化转型的企业，应当明确自身优劣势并做好阶段性规划、选择适合的路径，才能在激烈的竞争中获取优势。

关键词:企业数字化转型;忘却学习;利用式学习;探索式学习;K-means聚类分析;LightGBM算法

Abstract:

In the wave of digital economy, the replacement of knowledge such as technology and management models has made unlearning an effective way for enterprises to achieve digital transformation. However, relevant research lacks the identification of different enterprise characteristics and the unlearning methods adopted at different stages of digital transformation. Therefore, to provide targeted theoretical and practical references for digital transformation of enterprises, based on the perspective of unlearning, the degree of industry digitization, the level of regional digital economy development, and the age of enterprises are used as the basis for classifying enterprise types. The K-means algorithm is used to classify A-share listed enterprises in China that underwent digital transformation from 2017 to 2021 into three different groups: industry and regional win-win, industry advantage, and regional advantage. Strategy formulation, transformation implementation, and evaluation iteration are considered as the three stages of enterprise digital transformation. The lightweight gradient lifting machine algorithm (LightGBM) is used to obtain the learning methods of different types of enterprises in each stage of transformation and combine them into their digital transformation learning paths. The results show that when conducting digital transformation, industry and regional win-win enterprises prioritize the acquisition of external knowledge and gradually shift their focus to the abandonment of outdated knowledge. Industry advantageous enterprises focus on the acquisition of external knowledge throughout the process, while regional advantageous enterprises prioritize the abandonment of internal outdated knowledge and gradually shift their focus to the acquisition of external knowledge; In addition, digital transformation has a strong attraction for industry followers. Therefore, for enterprises planning digital transformation, they should clarify their own strengths and weaknesses, make phased planning, and choose suitable paths in order to gain advantages in fierce competition.

Key words:enterprise digital transformation; unlearning; exploitative learning; explorative learning; K-means clustering; LightGBM algorithm