We construct a multi-agent simulation on the business transaction model, and find the importance of third-party accreditation to keep the faithful transactions in the business world by the simulations.

米国発のサブプライム問題以降，株価下落を抑制するために各国の株式市場で空売り規制が発動された．しかし，この規制の効果については明らかになっていない．そこで本研究では，現実の金融市場を計算機上で模倣した市場（人工市場）を用いて，空売り規制が株式市場安定化にどのような効果をもたらすかを検証した．はじめに，空売り可能な人工市場と空売りが禁止された人工市場を構築し，株価変動を比較したところ，前者の方が株価が安定し，後者ではバブルが形成される可能性が高いことがわかった．次に取引途中に空売り規制が発動される場合とそうでない場合の株価推移を比較検討し，規制が発動される方が株価下落抑制力が高いことがわかった．以上より，空売り規制は，短期的には株価下落を抑制し市場安定化に貢献するが，長期にわたって規制するとバブルを形成し市場安定化の阻害になることが判明した．最後に理論株価が市場株価の影響を受けたときの収益率のボラティリティを観察し，市場株価の影響が強いほど市場が不安定になることを示した．

This research constructs the financial market model and analyzes the real market with evolutionary game theory. Especially, this research focuses on the order book in the financial market. This research is based on Kikkawa [7], extends that each player has a risk attitude and analyzes the order book with Micro-Econometrics's methods. We derive Nash equilibrium from this regression method and we present the method to forecast the next step.

2009年総選挙の結果を予測する予測市場実験shuugi.inの結果について報告する。

I review about speeding up of exchanges in Japan recently. This will trigger higher frequency trading, and investors will more need AI technologies.

We proposed an idea of order book analysis for predicting the direction of Nikkei 225 index in short terms using neural network. Not only the prediction of the market was proposed, but also the probability of execution was described as the first passage time of the Brownian motion in our model. Our system worked well for predicting the market behavior to some extent.

In this research, we describe the prediction of the stock price fluctuation by using Bayesian Network. Bayesian Network is trained with stock price fluctuations DJIA30 in New York stock exchange market, FTSE100 in London stock exchange market and NIKKEI225 in Tokyo stock exchange market. Then the network is applied to predict FTSE100 fluctuation. Firstly, FTSE100 fluctuation in 2007 is predicted by technical analysis and Bayesian Network analysis. The results show that the prediction accuracy of Bayesian Network is much better than that of technical analysis. Next, we will discuss the prediction accuracy of the Bayesian Network in 2007 (sub-prime loan problem). The results show that the prediction accuracy decreases not only at the time of the event but at the time of the policy change for the event.

This study attempts to uncover underlying information in stock message board (hereafter BBS) by using network analysis. It is said that many of postings on BBS could be noise. Therefore the overall sentiment of BBS often carries little useful content for future stock investment return. Moreover, under efficient market hypothesis (EMH), it is unlikely that investors (hereafter poster(s)) disseminate valuable information without compensation. However, some empirical research in the United States shows that there are a few posters in the community who post valuable information on BBS. The problem is how to extract such informed posters systematically from the BBS community. In this study, I utilize network analysis to solve this problem. Results of the empirical study on Yahoo! Stock BBS show that neither number of posting nor degree centrality could extract informed poster. However clique could extract informed poster. Return of informed posters is both statistically and economically significant even after risk adjusted. This indicates that the network analysis approach is valuable in screening out noise in BBS posting.

本論文では，複利リターンに基づく強化学習を用いて日本国債の取引戦略を獲得した結果を報告する．残存期間10 年の日本国債を対象として，複利リターンに基づくQ 学習と複利リターンに基づくOnPS を用いて5 年間の金利データから取引戦略を獲得し，その後の1 年間のデータを用いて評価した．

This paper reports our research in progress on relations between volume of transactions and related news articles. Our current goal is to make a system to predict vlolume of transactions of a brand from the news articles related to the brand. Our algorithm clusters news articles by LDA and predicts whether the volume of transactions at a target day will increase or decrease.