In this paper, we apply reinforcement learning algorithm to enhance return from technical trading in the Japanese stock market. Specifically, we employ MACD (Moving Average Conversion Diversion) signals to make buy / sell decision. MACD trading signal generates good return when the market is in a trending state but performs poorly when in a box-range state. Reinforcement learning endeavors to identify the state ex-ante and helps traders efficiently allocate capital. We demonstrate how we design our trading system and show the results of our simulations. Our results indicate the reinforcement learning of technical trading signals dramatically improves returns.
Agent programming models such as distributed multiagent models and mobile agent models have been proposed. We have proposed an agent programming model for business applica-tions since 2000 and applied to auction systems, a financial trading system, and others applications. Through those applications we obtained the knowledge that the programming model is eycient on developing applications. Especially, it is very eycient for scale out applications. In this paper, we discuss about the eyciency by referring a trading system developed on top of the model.
This study analyzes the sentiment of the Japanese economy that might appear in daily news articles. To quantify such a sentiment, we created an index that accounts for the frequency of occurrence of the words that affirmatively or negatively describe the current economic situation. Using articles taken from the Nikkei, we counted the numbers of "positive" as well as "negative" words in the articles. Constructing a daily summary index, we then conducted statistical analysis to examine correlations between the sentiment index and Tokyo Stock Exchange prices. One of our interesting findings is that the index significantly predicts stock prices of three day ahead.
In this paper, we analyze how the relationship among Nikkei average futures and individual stocks changes when a big event occurs, intending to give investors useful information for the risk management. We showed that while there were few particular relationships among Nikkei average futures and individual stocks before the Great East Japan Earthquake, just after that, the relationships greatly changed, and there were some trends peculiar to individual stocks in how the relationships changed.
This study summarizes some recent research topics on financial systems and systemic risk which were presented at Network Approaches for Interbank Markets, a workshop held in May at Catellon, Spain.
This paper proposes a method that extracts causal knowledge from Japanese financial articles concerning business performance of companies via clue expressions. Our method decides whether a sentence includes causal knowledge or not when the method extracts it. For example, a sentence fragment "World economy recession due to the subprime loan crisis ..." contains causal knowledge in which \World economy recession" is an effect phrase and \the subprime loan crisis" is its cause phrase. These relations are found by clue phrases,such as "ため(tame: because)" and "により(niyori: due to)". We found that some specific syntactic patterns are useful to improve accuracy of extracting causal knowledge. Therefore, our method can extract causal knowledge accurately.