| Summary: | 碩士 === 國立中正大學 === 資訊管理學系暨研究所 === 101 === When a group attempts to obtain a consensus in the decision-making process, because each participator has their self-awareness, it is liable to encounter the conflict problems, therefore, how to resolve conflicts to reach the consensus is the core of group decision-making. Recently, Chang proposed a new type of preference ranking model which can provide temporal relationships between items, it is based on sequencial pattern mining. However in real life, there are lots of interval-based circumstances, which can be described as the temporal relations of items more precisely, such as taking medicines, rental mechanisms or stock trends. Hence this study employs the concept of conflict operation of Chang to interval-based temporal pattern mining. In stock investment, time is a very important factor about whether investor can profit, and there are two investing preferences of stock investing strategy: short-term and long-term. When system collects investing preferences, there is a situation may occur is near number of participators give opposite opinions (short-term investment and long-term investment), for this reason, this study modified T-Apriori algorithm and added conflict operation to generate new algorithm, CITPM(Consensus Interval-based Temporal Pattern Mining). In experiment phase, this study designed a stock recommendation system to collect participators’ preference data and used CITPM to reach consensus interval-based temporal patterns. The result shown CITPM can not only generate traditional temporal pattern but also consensus temporal patterns by using conflict threshold. Furthermore, this study chose partial patterns for users to rate, and the result shows that the stock patterns recommended by a group of participators are worthy references for investors.
|
|---|
