Utilizing Technical Indexes and Database to Construct the Online Prediction System of Stocks

• Main Authors: Chia-Hua Chang, 張嘉樺
• Other Authors: 張百棧
• Format: Others
• Language: zh-TW
• Published: 2009
• Online Access: http://ndltd.ncl.edu.tw/handle/02333206479745035283

碩士 === 元智大學 === 工業工程與管理學系 === 97 === The main point and spirit of the technical analysis are to utilize statistical and mathematical formulas to discover, analyze and identify the rhythm and context of price fluctuation. In general, if investors can totally follow technical indicators such as the KD Index and the MACD Index to operate, they will absolutely enjoy a good performance in the long run. For the purpose to enable investors to make immediate decisions of investment when the buy signals appear, this research has utilized Dynamic Time Warping (DTW) and Piecewise Linear Representation (PLR) integrating Back-propagation Neural Network (BPN) to construct a DTW-PLR model and form a technical index system for trade decision which helps investors to detect the appropriate trade point and effectively reduce investment risk as well as to increase the profits. Back-propagation Neural Network is mainly to learn the connection weights between input variables and output variables; then Genetic Algorithms is applied to evolve better representation values which are expected to identify the better trade point in the future. If the technical index system is available at any time for the check of latest prediction indexes and the operation of database, the prediction can therefore provide the most accurate figures instead of frequently changing its conditions of basic analysis with the time-space environment. This change is quite slow; investors may suffer the extreme impact when not being able to endure temporary waiting. If the prediction figures can be immediately known, the analysis regarding unstable data can therefore be improved, and further to predict the new return ratio of investment. Although it is difficult to establish a highly-efficient technical index online prediction system, yet investors and users can obtain more latest information from the system which is developed by Java program and Oracle database.