The research theme of this project is cost reduction of Quarity Assurance(QA) for digital game. Focusing on QA dealing with ensuring game balance. In this research, we create player agents for the game by Genetic Algorithm and systems to support QA's work using statistics of agents' data. Agents behave like real players in eyes of developers of the game. Agents' data can provide us statistics regarding strength of various items.

These days in almost all manufactual companies in Japan, A lot of data are collected frequently everytime. These data has much useful information about the quality and operation of their production process lines. However, it is difficult for us to use the data well. these data has so many dimensions that it is hard to understand instantly what is wrong with their production processes and decide what to do. In this paper, I suggest a way to understand the condition of their production process lines easily with some knowledge of their masters to help us decide what to do to improve their quality or operation.

This paper show our research about treatment strategy of septicemia using DCP data with pLSA, probabilistic latent semantic analysis, which especially can model clients' change along with time. We used a variable which combine clients' ID and the days from hospitalization when we made some medical treatment and what we do as treatment to cluster our clients with pLSA, then summarized death rate and medical cost, days clients stay in this hospital of each cluster. Through researching clients' change along with time, We found some pattern how a client in some cluster move to another cluster those in this tend to die more, although there are some other clients who stay same cluster enough days after some treatments. Our research and method are accepted as some possibility of DCP data to assist treatment strategy in medical field.

Microscopic traffic simulations are useful for solving various traffic-related problems. An origindestination (OD) matrix is a typical representation form of traffic demands that is required for simulation. We have been proposed an indirect method to estimate the OD matrix using a traffic simulator as an internal model. However, due to shortage of complete real data, such as signals, estimated results often makes unrealistic traffic congestion in the simulaton. In this study, we introduce constraints to the optimization problem by comparing demand flow and performance flow. Through the experiments most of the unrealistic congestion cases are mitigated.