| Summary: | 碩士 === 國立臺灣科技大學 === 資訊工程系 === 97 === The main purpose of this thesis is to develop a steady Program Trading System of German Deutsche Actienindex (DAX) Index. Various collection of indicators are generated based on AiSM futures platform. By means of artificial intelligence, an effective program trading system is developed according to DAX index data by analyzing the properties of index fluctuation. Program Trading Models can be identified based on evaluation indicators. In this paper, selective models DAX_A to DAX_E have been proposed.
By analyzing the evaluation indicators, a single model is sensitive to a certain period of time which may not fit to the model. Such sensitivity may cause great loss and generate high pressure to the model users. Therefore, we proposed a mean-oriented multi-model trader to diversify the risk with the potential of extra cost.
In order to overcome the disadvantage of mean-oriented method, we further propose a Mapping Method which is able to generate reasonable mapping tables by applying statistical analysis on mean-oriented method. By combining both models, we will be able to ignore inefficient portion, highlight the portion with high efficiency, and achieve the goal of reducing cost.
|
|---|
