Published papers

本研究は、ChatGPTの登場が日本株式市場に与える影響について分析を行う。分析対象は、国内上場企業とし、分析期間は、ChatGPTが公表された前後の期間とする。本分析では、株式価格変化を基に影響の評価を行うが、株式分析には、双方向LSTM-GNNを用いる。また、結果の解釈にXAIを用いることで、分析結果の可視化を試みる。

Corporate valuation is one of the important financial modeling in IPO and M & A. In particular, at the time of IPO, there is no underlying stock price information, so its price depends on the result of corporate valuation. However, the offer price at the time of the IPO often has an under-pricing in which the public price falls below the initial value, resulting in an opportunity loss for the issuer company. Although many previous studies have referred to this issue, no consensus has yet been obtained. In this study, we consider that there is room for improvement in the corporate value evaluation model and try to evaluate the corporate value by reproducing the initial value of a newly listed company using patent information. Specifically, first, Derwent World Patents Index held by the company is extracted, and the patent document is vectorized by using Sparse Composite Document Vector. Next, the center of gravity of the patent is calculated from the vectorized patent document, and the similarity between the companies is expressed by the distance to select similar companies. Finally, using the selected companies, a comparable peer company multiple method is performed to evaluate the validity of the stock price. In conclusion, it is suggested that the model is more explanatory than using industry average financial figures. In this study, we analyzed feasibility of our proposed method.

情報技術の発達に伴い，株式市場分析においても情報の重要性が注目されてきている．本研究は，ニュース記事に着目し，情報技術を用いて投資家が判断や意思決定を支援するモデルの構築を目指す．その方法として，TOPIX Core30 の構成銘柄のニュース記事の要約文章を生成し，分散表現に変換して機械学習で分析を行った．

Stock market fluctuations are caused by various factors. This study focuses on news articles toimprove the accuracy of stock market predictions. The effect of news articles is bigger when informationthat was not predicted is public, earning surprise. We have summarized past news articles by GPT-2 todiscover news trends. We also converted the news articles into vectors and link them to stock prices bymachine learning. As a result of this analysis, we were able to improve the accuracy of stock market pre-dictions.

Japan is suffering from a serious labor shortage due to the economic expansion anddemographic change in the non-regular employment of the labor market. In particular, non-regularemployees account for 80% of employees in the food-service industry is a critical issue. The purpose ofthe present study is to analyze a mechanism of job matching focusing on non-regular employmentthrough agent-based modeling.