【Session Theme: Other AI x OS Topics】
Candlesticks are graphical representations of price movements for a given period. The traders can discover the trend of the asset by looking at the candlestick patterns. Although deep convolutional neural networks have achieved great success for recognizing the candlestick patterns, their reasoning hides inside a black box. The traders cannot make sure what the model has learned. In this contribution, we provide a framework which is to explain the reasoning of the learned model determining the specific candlestick patterns of time series. Based on the local search adversarial attacks, we show that the learned model perceives the pattern of the candlesticks in a way similar to the human trader.
Our approach produces two contributions. The first is that our GAFCNN model constructs an innovation field of financial vision research for candlestick recognition. The second is that we propose an approach based on the modified local search adversarial attack to explain the reason for the GAF-CNN model on how to determine the different candlestick patterns. Our GAF-CNN model can identify eight types of the candlestick and understands the feeling as a human has seen. We can confirm that the GAF-CNN model has indeed learned the sense of the candlestick from the trader. The GAF-CNN will be perfect for building a complete explainable trading model. We provide an open-source implementation and training data for the paper in the following URL: https://github.com/pecu/FinancialVision.
About PecuLab (peculiar change the world)
Yun-Cheng Tsai (蔡芸琤) is the PecuLab funder and an assistant professor in the School of Big Data Management at Soochow University. PecuLab creates a new research field - Financial Vision until now from 2018 summer. She had working experience in research and development of telecommunications system company for four years before entering the Ph.D. program. Then, Yun-Cheng had an additional two years of experience as a bank consultant and more than five years of teaching programming experience in the Information System Training Program at National Taiwan University during the Ph.D. program. She graduated from the Department of Computer Science and Information Engineering at the National Taiwan University. She started to work for her alma mater in the Center for General Education from 2016 summer to 2019 summer. Her responsibility was designing and teaching interdisciplinary computational thinking courses, including data science, data visualization, and data analysis. It’s a new term CS+X, which means Computer Science for anything. The working experience cultivates her strength of interdisciplinary teaching and research.