Katsuari Kamei

Borderline over-sampling for imbalanced data classification

Traditional classification algorithms usually provide poor accuracy on the prediction of the minority class of imbalanced data sets. This paper proposes a new method for dealing with imbalanced data sets by over-sampling the borderline minority class instances. A Support Vector Machine (SVM) classifier is then trained to predict future instances. Compared with other over-sampling methods, the proposed method focuses only on the minority class instances residing along the decision boundary, due to the fact that this region is the most crucial for establishing the decision boundary. Furthermore, the artificial minority instances are generated in such a way that the regions of the minority class with fewer majority class instances would be expanded by extrapolation, otherwise the current boundary of the minority class would be consolidated by interpolation. Experimental results show that the proposed method achieves a better performance than other over-sampling methods.

Online learning from imbalanced data streams

Learning from imbalanced data has conventionally been conducted on stationary data sets. Recently, there have been several methods proposed for mining imbalanced data streams, in which training data is read in consecutive data chunks. Each data chunk is considered as a conventional imbalanced data set, making it easy to apply sampling methods to balance data chunks. However, one drawback of chunk-based learning methods is that the update of classification models is delayed until a full data chunk is received. Therefore, this paper proposes a new method for online learning from imbalanced data streams, which uses naive Bayes as the base learner. To deal with the problem of class imbalance, a new training instance from the minority class is always involved in learning, but one from the majority class is only used with a small probability. In effect, this method corresponds to an under-sampling technique on imbalanced data streams. We show the effectiveness of the proposed online learning method on ten UCI data sets of various domains. Problems in the performance of naive Bayes on imbalanced data sets are also discussed.

Human Reasoning Awareness Quantified by Self-Organizing Map Using Collaborative Decision Making for Multiple Investment Models

Collaborative Decision Making (CDM) is one of the concepts of human reasoning awareness, which refers to expert knowledge of the group and its preferences in a dynamic market environment. In this paper, we present a new approach, which is a framework for collaborative decision making, together with expert feelings about market dynamics to deal with multiple models of stock investment portfolios. The framework aims to aggregate collective expert preferences, including of group expert psychology and sensibility, assists a dynamic trading support system and achieve the greatest investment returns. Kansei evaluation uses to quantify trader sensibilities about trading decisions, market conditions with uncertain risks. Collective group psychology and preference of traders are quantified that represent in membership weights. The framework is used to quantify Kansei, quantitative and qualitative data sets, which are visualized by Self-Organizing Map (SOM) in order to select the best alternatives with dynamic solutions for investment. To confirm the models performance, the proposed approach has been tested and performed well in stock trading for stock investment portfolios. The experiments through case studies show that the new approach, applying Kansei evaluation enhances the capability of investment returns and reduce losses to deal with various financial investment models.

Proposal of Group Decision Support System Using SOM for Purchase of Automobiles

This paper describes a group decision support system (a negotiation support system) based on Kansei of group members using SOM for automobile purchase for family. Firstly, we evaluate Kansei scores of automobiles existed in real world using Kansei engineering. Secondly, we make a SOM map based on both the Kansei scores of the real automobiles and ideal automobiles of the group members. If Kansei scores of the ideal automobiles of members are far away from each other on the SOM map, the system shows some compromise proposals to the members for negotiation. Where, the compromise proposals are made by small changes of Kansei scores of members. Finally, the Kansei scores of their ideal automobiles gather into a small area on the SOM map, then the system decides some real automobiles nearest to the area as an negotiation result of members.

Proposal of group decision support system for commodity purchase using Kansei and SOM

This paper describes a group decision support system (a negotiation support system) based on Kansei of group members using SOM. Firstly, we evaluate Kansei scores of commodities existed in the real world using Kansei Engineering. Secondly, we make a SOM map based on both the Kansei scores of the real commodities and ideal commodities of the group members. If Kansei scores of the ideal commodities of members are far away from each other on the SOM map, the system shows some compromise proposals to each member for negotiation. Where, the compromise proposals are made by small changes of Kansei scores of members. Thirdly, the Kansei scores of their ideal commodities gather into a small area on the SOM map, then the system decides some real commodities nearest to the area as a negotiation result of members. Finally, we show an usefulness of the proposed system using satisfaction rates of the members to output results of the system.

A Coevolutionary System for Development of Strategies in Poker Game

We propose a coevolutionary system which reciprocally develops players strategies in two player games. The game environment is the seven stud poker that is a complex real- world game of imperfect information. In our system, the players decide their actions based on a self-learning by Classifier Systems and then make the strategies more complex and excellent. We analyze dynamics of the evolution of the players strategies and show the learning process of reciprocating skills of players.