Adaptive Quantitative Trading: An Imitative Deep Reinforcement Learning Approach
Authors: Yang Liu, Qi Liu, Hongke Zhao, Zhen Pan, Chuanren Liu2128-2135
AAAI 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Experimental results demonstrate that our model can extract robust market features and be adaptive in different markets. Experiments We back-test our model on the minute-frequent futures data with practical constraints. |
| Researcher Affiliation | Academia | Yang Liu,1 Qi Liu,1 Hongke Zhao,2 Zhen Pan,1 Chuanren Liu3 1Anhui Province Key Laboratory of Big Data Analysis and Application, School of Data Science & School of Computer Science and Technology, University of Science and Technology of China 2College of Management and Economics, Tianjin University 3Department of Business Analytics and Statistics, University of Tennessee {liuyang0, pzhen}@mail.ustc.edu.cn, qiliuql@ustc.edu.cn, hongke@tju.edu.cn, cliu89@utk.edu |
| Pseudocode | No | The paper describes the proposed algorithms (Recurrent Deterministic Policy Gradient, Demonstration Buffer, Behavior Cloning) using prose and mathematical equations, but it does not include a formally labeled 'Pseudocode' or 'Algorithm' block. |
| Open Source Code | No | The paper does not contain an explicit statement or link indicating that the source code for the described methodology is publicly available. |
| Open Datasets | No | We collect minute frequent futures data from Join Quant.com 6, a famous Chinese quantitative platform. (Footnote 6: https://www.joinquant.com/) While a platform is mentioned, a specific dataset and a direct link/citation for public access to the *exact* dataset used are not provided. |
| Dataset Splits | No | The Data in the training set spans from Jan 1st, 2016 to May 8th, 2018 while the data in the testing set spans from May 9th, 2018 to May 8th, 2019. The paper specifies training and testing sets, but does not explicitly mention a validation set or a three-way split. |
| Hardware Specification | No | The paper mentions that 'With the assistance of computer technology' quantitative traders aggregate information, but it does not provide any specific details about the hardware (e.g., CPU, GPU models, memory, or cloud resources) used for running the experiments. |
| Software Dependencies | No | The paper discusses various techniques and models like Recurrent Deterministic Policy Gradient (RDPG) and Gate Recurrent Unit (GRU), but it does not specify any software dependencies with version numbers (e.g., Python 3.x, TensorFlow 2.x, PyTorch 1.x) that are required to replicate the experiments. |
| Experiment Setup | Yes | For better simulation, we take into accounts practical constraints, including the transaction fee δ = 2.3 10 5 and the constant slippage ζ = 0.2. Furthermore, we assume that each order can be traded in the opening time of each minute bar and the reward is calculated in the closing time. Additionally, exclusive features about the futures including margin, contrast settlement are considered. The training epoch will be broken off once positions are lost by 50% or lacking margin. We initialize our account with $ 500,000 in cash at the beginning of the test. |