@inproceedings{wang2019conservative,
author = {Wang, Lewen and Liu, Weiqing and Yang, Xiao and Bian, Jiang},
title = {Conservative or Aggressive? Confidence-Aware Dynamic Portfolio Construction},
booktitle = {2019 IEEE Global Conference on Signal and Information Processing (GlobalSIP)},
year = {2019},
month = {November},
abstract = {Indicator-based investing is a popular investment strategy driven by technical analysis for the stock market, the key issue of which is to construct portfolios from technical indicators. Due to the high volatility and non-stationary of the stock market, the effectiveness of an indicator, however, varies largely across different periods, which has made it necessary to dynamically adjust indicator-based investing. In this paper, we propose a confidence-based calibration approach for dynamic portfolio construction. The major intuition behind is to tune a more concentrated portfolio when the indicator yields higher confidence otherwise a relatively equal-weighted one. To seek a maximized long-term profit, we further propose to integrate learning the confidence (i.e., future effectiveness) of an indicator into a unified portfolio construction approach powered by a recurrent reinforcement learning framework. Compared with the traditional indicator investing strategies, our confidence-based calibrated indicator of investing can obtain significantly higher returns with lower risks.},
url = {http://approjects.co.za/?big=en-us/research/publication/conservative-or-aggressive-confidence-aware-dynamic-portfolio-construction/},
}