Profit prediction based on management discussion and analysis of annual report of listed companies

Non-financial data has gradually become important indicators for investors to measure the value of enterprises. China Securities Regulatory Commission also requires listed companies to transmit reliable information by disclosing business information. Therefore, this paper introduces tone research, an important feature of linguistics, into the fourth chapter of the annual report, management discussion and analysis (MD&A). By dictionary building, Jieba segmentation and snowNLP sentiment analysis, it preprocesses and quantifies the tone of 1005 annual reports in 2016-2018. And it establishes a model to study the relationship between tone and the past year's business situation and the next year's performance, and analyzes the prediction function of tone. The results show that if the listed companies suffer from operating losses, the tone of MD&A in the annual report is more negative. The tone of MD&A has obvious information increment and can provide predictions for investors. The positive tone of this year's annual report text is positively related to the future profitability index. In addition, the companies with high business risk coefficient will positively affect the positive tone of MD&A. The results of this paper summarized the MD&A tone analysis and put forward some suggestions for information users.