| Summary: | 碩士 === 輔仁大學 === 統計資訊學系應用統計碩士班 === 107 === Nowadays in Taiwan, much of the population’s monthly income barely covers daily expenses. In order to improve the quality of life, people try to find investments that can increase their passive income, which include temporary and long-term stock investments. According to statistics, the Taiwan stock market issues more than NT$1 trillion in cash dividends per year. Therefore, this makes investing an effective method to increase net worth.
The present study utilizes TEFI imaging, Convolutional Neural Networks, Pre-Trained Models, and Long Short-Term Memory (LSTM) that were formed by Intraday Data for the Taiwan Stock, Time Series Feature Extraction, and Technical indicators to establish Real-Time Time Series Evolutionary Feature Image (TEFI) and discuss the effectiveness of model predictions.
The results reveal that by using a Pre-Trained Model to extract features from TEFI imaging and combining with other Deep Learning Models, the results are all better than the general Deep Learning Model, which does not use TEFI imaging. As a result, we now know that TEFI imaging provides abundant information and Pre-Trained Models are able to extract features from TEFI imaging.
|
|---|
