Hirotaka Mizuno

Applications of fuzzy logic/neural network to securities trading decision support system

The effectiveness of applying fuzzy logic and neural networks to securities trading decision support systems (STDSS) is demonstrated through some examples. First, the characteristics of STDSS are discussed. Next, examples such as buy/sell timing detection or stock portfolio selection using fuzzy logic and neural networks are reported by showing their algorithms and simulation results. In a test of the models ability to follow price trends, the model correctly identified 65% of all price turning points.<<ETX>>

Remote consultation system for participants with different skills and roles

In this paper, we propose a remote consultation system in financial and public areas, which allows customer and consulting staff to communicate with the same productivity as in face-to-face consultation. For remote consultation, current groupware technology alone is not enough since it is intended to support participants who are uniform in the sense of sharing same objectives, roles and knowledge, and having same skills of computer operation. Basic design concept of the proposed system involves supporting remote conversation among participants with different skills and roles. For this purpose, we propose following additional support functions for these two types of participants to operate terminals and communicate smoothly: 1) tele-monitoring of customer and customers terminal; and 2) tele-operating of customers terminal. Experimental result of user testing and practical interaction data demonstrate that the proposed functions are powerful to support smooth remote consultation.

On extraction of WWW pages with spatial information for mediating their distribution

Maps are expected to play an important role that interfaces help users search on the WWW. There is much area-based information on the WWW, and there are also many service providers that support such information. The implementation of spatial information sharing services is expected to mediate these contents and services. In developing information mediation services, preparing rich contents for the service is one of the main problems. We implemented an agent, that extracts spatial information and carried out an experiment to extracting spatial information automatically using this agent. The agent extracted 23% of eleven thousand two hundred WWW pages using text address information, which is one type of spatial information. We confirmed the accuracy of the extraction manually, and found there was no noise in this case. This result shows the agent can accurately extract text address information, and is therefore very useful.

Map-based information mediation on the web with float coordinate system

Recently maps have been required to play an important role in the user interfaces of information discovery from the WWW. For the further implementation of spatial information sharing service. We adopt a map-based mediation architecture, in where are three roles of contents provider, mediator and requester, to integrate the Web contents and their services. In this paper, we propose the basic concept of the service and an idea to solve its technical problem. Significant challenge in this topic is the establishment of interoperability among a variety of contents providers. Aiming at strict data conversion, we propose hyper structural location management with float coordinate system, which defines the location of a content in relativistic location of specific map. By following the hyperlinked relationship, registered map should be found in specific location with its base coordinate system. In addition, hand-written deformed map can be managed in the same way. We designed the basic specification of spatial information sharing system with float coordinate system and developed the application as a mediator system.

Data queries using data visualization techniques

Proposes a flexible and friendly user interface for data queries using data visualization. There are a variety of ways how computers help users search data, and query interfaces are expected to allow users to choose these ways according to their purposes. These ways are classified into two aspects, filtering and browsing. While filtering is a function which searches data on behalf of users, browsing is function which helps users to search data. In the proposed query interface, both browsing function and filtering function are integrated as follows. The browsing function consists of data visualizing and image zooming. With data visualizing, the original data can be transformed into representations from views, which helps users understand the data more easily. Then the image zooming function displays the visualized representation in various scale factors, which enables users to investigate the outline of the entire data or the detail of partial data. The filtering function is also enhanced by providing several ways of post-processing the filtered results. These include marking the results on the visualized representation as well as extracting them. We implemented an experimental system using the proposed query interface. The system shows that it works well as a post-processing unit to conventional data query interfaces, and it is powerful for flexible handling of the data which are roughly searched with the conventional query interfaces.

Visual decision support system

This paper describes a visual decision support system that supports marketers in the advertising field. The purpose of this system is to help users estimate the effect of advertising based on the extraction of expert knowledge when there is no fresh data to go on. This system includes the following functions: a) To check the dispersion of data and the behavior of models using 3D plot displays, and b) To make prediction models capable of estimating the effect of advertising using a constrained learning algorithm for neural networks. These functions have each of the following features: a) The ability to easily check the dispersion of data and the behavior of models in the various axes, allowing a marketer to delete unnecessary data in 3D spaces. b) The basic premise that the relationship between the input and output monotonously increases. This system is useful for domains where only deficient data is available like estimating the effect of advertising. Experimental simulations using real data have shown that experts can easily check the dispersion of data and the behavior of models, and have shown that the algorithm makes more reasonable models for estimating the effect of advertising than other conventional algorithms.

Intelligent guidance method for remote consultation systems using history of user's reference

Recently, remote consulting service systems along the network are developed. This paper describes the method to guide a user of such a system to the appropriate consultant automatically. The method, firstly, analyzes the history of users references to the information server provided by the system and infers the users needs from the references. Secondly, it assigns an appropriate consultant according to the extracted needs. The needs extraction method is evaluated through an experiment and verified its power of extraction. We also describe some methods to improve the precision of the needs extraction and compare them with the original method.

Computer supported tele-consultation system under cooperative work among customers, branch office workers and expert staffs

In the field of financial services, service improvements are needed for various kinds of customer requests and for effective utilization by bank workers. A teleconsultation system using multimedia and network technologies is assumed to be effective, and has come into practical use, but in the case of using a distributed terminal without any operating staff, it is difficult to provide effective services because customers are unfamiliar with the terminal operations, and the number of connecting expert staff members is limited. This paper proposes a method of supporting a teleconsultation with a concentrated communication control facility among a tripartite composed of a customer, a clerk at the banks branch office and an expert member of staff at the banks head office, who has the consulting task.

Constrained neural network learning algorithm for estimating advertising effectiveness

This paper proposes a neural network algorithm for estimating advertising effectiveness. The basic premise under this algorithm is that the relationship between input and output monotonically increases. The algorithm is useful for domains such as estimation of advertising effectiveness, for which only deficient data are available. The experimental simulation shows that the algorithm builds a reasonable model for estimating advertising effectiveness.

Interactive Consultation System with Asymmetrical Communication between People in Different Electronic Communities

Recently, in electronic community, information retrieval and communication have bcome popular among people with different cultures, variety of circumstances and organization in electronic communities. However, it is sometimes difficult to find suitable information and services in a short enough time for real use, because quality of information on internet is not unified and quantity of information is enormous on internet. In this paper, we discuss communication issues in electronic communities, especially asymmetrical communication between experienced experts and residents with less information. These are two view points: 1) Retrieving Heterogeneous Information Sources, and 2) Integrating Heterogeneous Information Sources.