基于自然语言处理和机器学习的长期股权投资分类模型

Long Term Equity Investment Classification Model by Using Natural Language Processing and Machine Learning

长期股权投资作为投资活动的重要类别,对其进行正确分类并使用恰当的会计核算方法十分重要。以往因投资项目内容的描述均为文本,传统基于数值的计算机分类模型难以直接实现分类。本文尝试通过自然语言处理和机器学习技术,对文本数据进行分词处理后,使用TF-IDF统计方法将文本转为数据向量,并利用人工标注过分类的训练数据构建支持向量机模型,实现对长期股权投资项目的智能分类。模型对上市公司长期股权投资项目的分类准确率可达71%。本文的研究结果表明,人工智能方法可以辅助财务人员进行会计业务分类判断。

Long-term equity investments are important parts of a firm’s investment activities. It’s critical to do a correct classification and use an appropriate accounting method on long-term equity investment. Because description about long-term equity investment is mostly recorded in text form, traditional asset classification models can not directly use textual data as input. In this paper, we apply natural language processing method and machine learning method to construct a classification model. Specifically, we use Jieba to do word segmentation and use TF-IDF method to get the vectors of textual data. We then construct the Support Vector Machine based on the vectors of the training dataset which has been labelled, and select different kernel functions and adjusting parameters to get the optimization model. The accuracy of the model is 0.71, indicating that the model can be used on the primary stage of the classification on the longterm equity investment. Our finding also verify the feasibility of applying machine learning to accounting transaction processing.