In stock markets, it is often argued that increased liquidity contributes to the public benefit of the market as a whole, but it is not self-evident. In this study, we analyze the impact of increasing market liquidity on traders' utility mathematically. We calculate an exact solution of an average expected utility for one trader by using a simple model in which we assume orders follow independent uniform distributions. However, even when we assume a more complicated model where orders interact with each other, we obtain the result consistent with the first simple model in the limit of infinite number of orders. Also, we define a balance price on the order book, and we analyze the behavior of the balance price.

In addition to the need for electricity consumers to take into account the complex behavior of the electricity market as a result of electricity deregulation, CO2 emissions from economic activities have also become an issue in response to recent calls for decarbonization, making the electricity procurement environment increasingly complex. In this study, we focus on electricity procurement by a factory as a large consumer in such a complicated electricity sector. We have conducted simulation experiments and evaluations using an electricity market multi-agent model for several scenarios, focusing on the benefits to the demand side of demand response (DR), which is attracting attention as a means of stabilizing supply and demand in the electricity system, in terms of cost and CO2 emission reduction effects. The results show the effectiveness of DR that takes into account the characteristics of the season and time of day, and the effectiveness of demand shifting that utilizes out-of-service hours.

Financial market microstructure has been scrutinised by utilising the high-frequency order-book data these days. In this talk, we report our preliminary data-analytical results on the persistence of the order flows (called the long-range correlation (LRC) in the literature) from microscopic dynamics. Empirically, the order flow is known to exhibit the persistence: i.e., the future order sign is strongly correlated with the historical sign sequence for a long time. This intrinstic character of financial markets have been a debatable issue in terms of its microscopic origin. One of the microscopic hypotheses to explain this LRC is the ordersplitting behaviour at the level of individual traders. We have carefully tested this microscopic hypothesis through microscopic data analysis of a large dataset in the Tokyo Stock Exchange, particularly from the viewpoint of the direct validation of the Lillo-Mike-Farmer model.

Deep Hedging, which uses deep learning and price time-series simulations to optimize option hedging, has recently been in the spotlight because it enables more realistic hedging that can take into account frictions such as transaction fees (imperfect market). However, the situation of hedging an option by other options has never been addressed by deep hedging because of its simulation difficulties. In that situation, pricing for tradable options should also be performed via deep hedging in simulations for realizing imperfect market simulations, which has required unrealizable enormous computational resources because of the nested architecture of deep hedging. Thus, in this study, we proposed a new deep-hedging mechanism for learning hedging strategies under such a nested situation. As a result, we showed better hedging via proposed deep hedging with multiple tradable options.

In recent years, research on stock trading and foreign exchange trading using technical analysis has been vigorously conducted. In the research on investment strategies using technical analysis, it is popular to construct trading strategies using deep reinforcement learning and neural networks. However, trading strategies constructed by these methods cannot be interpreted because they are not algorithms that take interpretability into account. Therefore, it is difficult to analyze the reasons for the actual trades. In this study, we propose a new algorithm, Weighted Genetic Network Programming, which is an improvement of Full Range Genetic Network Programming, one of the evolutionary computation methods. We propose a new algorithm, Weighted Genetic Network Programming, which is a modification of Full Range Genetic Network Programming, one of the evolutionary computation methods.

In recent years, research on financial transactions using deep reinforcement learning, which is one of machine learning, has been actively conducted. In these studies, various approaches have been taken, such as those that consider the number of financial instruments bought and sold, compound interest calculation, and those that use stock price charts for input, but enough profit has not been made in all periods. It is considered that this is because the opportunity loss cannot be taken into consideration. Therefore, in this study, we build a model to learn the optimal buying and selling timings to make a profit in stock investment by incorporating the opportunity loss for each action into the reward in deep reinforcement learning, and show its effectiveness.

In the past few years, there has been a growing interest in cryptocurrencies. However, the risk of incurring losses is high due to the large price fluctuations. Therefore, we want to reduce this risk by predicting the rise and fall of these prices. In this study, we use a convolutional neural network model trained on candlestick charts to make price predictions. In this experiment, the system was trained on the image pattern data of a set of five candlesticks, and predicted whether the price would go up or down. The model trained on the data from 1-minute intervals gave the best prediction accuracy.

For traders, it is important to minimize execution costs and achieve more efficient order execution. Since the mechanisms for incurring costs are unclear, being able to properly account for them will lead to lower execution costs and higher revenues. In order to achieve order execution with minimal costs, methods that model and infer market principles have been used. In recent years, model-free offline reinforcement learning methods have widely been utilized. However, the data on financial instruments contains a lot of noise, which makes learning hard and makes it difficult to converge to the optimal trading method. In this paper, we propose an optimal order execution method that improves performance by imposing constraints on the model. Through experiments, we have found that by imposing appropriate constraints, we can improve the performance of the optimal order execution method. We show that by setting appropriate constraints, we can achieve improved order execution compared to conventional methods.

There are three main factors known to exist in the foreign exchange market: carry, value, and trend, but the effectiveness of these factors is temporally changed by global market conditions. In this research, we extract latent factors from foreign exchange rates by a data-driven approach using the autoencoder. Then, we evaluate the possibility of market mispricing by the concept of anomaly detection, considering the exchange rate restored by the autoencoder as the theoretical value expressed by the factor-risk premiums. Moreover, assuming that the misprice is immediately modified into the reasonable value, we construct a portfolio of multiple currency pairs based on the modification process of mispricing. As a result, we confirmed the possibility of obtaining the excess return (i.e., alpha) as in the stock portfolio management.

In the EC logistics business, which ships products upon receiving orders from customers via the Internet, determining the products to be handled and their inclusion ratios are a fundamental problem to be solved to increase the expected profit. It is difficult to find a good solution in a finite time because the number of products handled in EC is more than several million and the number of combinations increases explosively in the order of nth power. We formulated this problem as a portfolio optimization problem and tested whether it is possible to find an optimal or good solution using Markowitz's mean-variance model. The motivation for applying the mean-variance model, which is usually used for diversified investment in stocks, to EC logistics is that the model is very simple because it is obtained by the expected value and its variance, and it is easy to apply to the modeling of indicators such as sales and inventory cost in EC logistics. In logistics, it is easier to control inventory management and logistics costs if demand fluctuations are mitigated and leveled out as much as possible. The mean-variance model, which considers fluctuations as risks, is easy to apply to EC logistics because fluctuations tend to lead to higher costs. In this study, we constructed a mean-variance model of sales and inventory turnover using actual EC logistics data, and confirmed the effectiveness of the method for determining a preferred product portfolio from both EC and logistics perspectives.

In this report, we attempt to evaluate the content of Japanese integrated reports issued by Japanese companies by textual analysis. We propose a simple evaluation model for Japanese integrated reports based on word2vec. Based on the observation of evaluation scores by the model, it is confirmed that the average quality of contents of Japanese integrated reports has been improving year by year. Furthermore, the level and the difference from the previous year of the scores show consistent relationship with the results of existing external evaluations by professional investors.

Given the voluntary nature of environmental, social, and governance (ESG)-related information disclosure in Japan, we use a sample of TOPIX firms from 2011 to 2019 to examine the relevance of internal and external corporate governance factors and ESG-related information disclosure for Japanese companies. For the internal governance factors, the results of the logistic regression analysis show that variables such as board independence and board activity significantly influence a company's ESG disclosure strategy, whereas the results of the generalized additive 2 model show that the internal governance variables are relatively less important than the external governance variables such as share ownership structure. For the external governance factors, the logistic regression results show that all the explanatory variables are significant. Although the results from the generalized additive 2 model are generally similar, non-linear relationships for institutional ownership and the Government Pension Investment Fund are also found. These empirical results suggest that the development of corporate governance frameworks such as Japan's Stewardship Code and Corporate Governance Code influences firms' ESG disclosure strategy and encourages them to disclose ESG-related information. This study provides new insights into the relationship between corporate governance and ESG-related information disclosure practices in Japanese companies.

Regarding ESG investments, which is one of the investment methods to evaluates cooperate environmental efforts, are becoming the minimum requirement for business. Information disclosure regarding Carbon Net-Zero society has led to the trust of the market and has begun to be directly linked to corporate value. Therefore, in this study, we visualized the relative progress of each decarbonized topic in each country. Specifically, we extracted articles related to decarbonization from about 40 million articles over the past three years, clustered BERT embedding vectors for those articles, and extracted characteristic words for each cluster by using c-TF-IDF. From this study, we found that decarbonization efforts are contrasting in Europe, the United States, China, and Japan. In Japan, we found that although the fields of "fuel ammonia" and "carbon dioxide capture and storage (CCS)" are attracting attention internationally and are leading in technology, they are inferior to other countries in terms of policy and regulation.

Fund managers at investment trust management companies make investment policy decisions by referring to the results of research on candidate companies for investment compiled by analysts. However, when there are many candidate companies, it is necessary to refer to a considerable number of reports, which is considered dificult to read carefully. Therefore, technology is required to (1) accurately determine the business sentiment of the companies concerned and (2) extract important information for investment decisions from the contents of the reports. In this study, we developed a machine learning model that predicts the rating, which is a rating index for investment decisions, in order to support the work of fund managers, especially for the requirement (1). There are two types of ratings that afiect investment decisions: outperform and underperform. Since the number of cases that fall into these two categories is small compared to other ratings, we attempted to expand the data of documents that give the same rating. As an existing data expansion method, there is a method to expand data by synonyms that can be obtained from WordNet. In this study, we propose a method to expand data based on the frequency of occurrence in financial documents. As a result of experiments, we verified the efiectiveness of the proposed data expansion method.