신경망과 의사결정트리를 이용한 Stream Data 예측 시스템 설계 및 구현

Abstract

At stock market, about stock price index prediction the research was attempted to use many techniques from various field. Generally the stock price index prediction is using the fundamental analysis which uses the immanence value of technical analysis and stock market. But, stock market of fact about reduction prediction or estimate of situation the description below is difficult with complexity and uncertainty of stock price index. In this research, In order to raise effectiveness of stock price prediction system uses the neural network and a decision-making tree algorithm. because Information which are useful is latent below-mentioned mining. In order accomplishes a real-time data mining effectively from prepares the data which is appropriate probably has the necessity to which what kind of data becomes mining. It processes a data with the form which is suitable at mining and will use at mining the technique which and selects the result which has become mining and the chain which applies in prediction interprets process in necessity. Also data mining relationship of the patterns which are useful is immanent in data or variables between uses the analytical model which is elaborate and seeks and accomplishes the role which reconfirms the experience knowledge of existing with the work and does not recognize simultaneously until now, provides new information and is a system which predicts. Consequently, this research uses the neural network and a decision-making tree and predicts the hereafter (tomorrow) rise and a depreciation of KOSPI 200 quotients the dual classification (binary classification) model which the prediction efficiency of each classification models and it plans it compares. The different meaning goal uses the prediction model which finally is constructed and to embody the prediction system of the individual stocks from stocks spot market, the contents which is concrete with afterwords is same. 1. Embodies the dual classification system which predicts the rise and a depreciation of parting stock price index from stocks spot market. 2, Uses a decision tree algorithm about the data which has a time series quality and reveals the prediction system which is efficient embodies the union or a relation of data and between. 3. After constructing the neural network and a decision-making tree model, divides TEST1 and TEST2 duration and simulation leads and the system which raises the accuracy of prediction about future point of view embodies. 4. Standardization process and above removal leads and application embodies the system which is possible from the actual money market.