| Summary: | 碩士 === 國立臺北科技大學 === 資訊與財金管理系 === 106 === In recent years, with the accumulation of open data and the progress of data mining methods, and most of the information on the network is mainly unstructured, the technique of text-mining helps readers quickly find and analyze the most important information, which is considered an important field. In the field of financial analysis, most of the past studies used financial data as the basis for analyzing the company’s stock price and the future development. With the development of technology and the improvement of the efficiency of processing high-dimensional data, data mining is applied to financial statement. The unstructured data of the report, combined with data-type data and non-structured text data, will reveal more discoveries than the data.
Our study uses the 10-K financial report to classify topic models by machine learning, and discusses whether the managements statements and future development descriptions in financial statements describe the companys stock price performance. We observe there is a negative impact on the return if it appeared in the financial report as a subject under the topics of future development; if a consumer-oriented topic emerged in the financial report, it would have a positive impact on return.
We propose a different way from the past investors to predict the performance of stock price by means of data, and provides knowledge discovery of the financial report text. It also provides investors with a new tool choice for the pre-stage of asset allocation.
|
|---|
