The Kelly Growth Optimal Portfolio with Ensemble Learning
Authors: Weiwei Shen, Bin Wang, Jian Pu, Jun Wang1134-1141
AAAI 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We analyze the behavior and hyperparameter selection of the proposed strategy by simulation, and then corroborate its effectiveness by comparing its out-of-sample performance with those of 10 competing strategies on four datasets. Experimental results lucidly confirm that the new strategy has superiority in extensive evaluation criteria. |
| Researcher Affiliation | Collaboration | 1Shanghai Key Lab for Trustworthy Computing, School of Computer Science and Software Engineering East China Normal University, Shanghai, China 2GE Global Research Center, Niskayuna, NY, USA 3Credit and Portfolio Risk Management, Citigroup, St. Louis, MO, USA |
| Pseudocode | Yes | Algorithm 1 Ensemble Growth Optimal Portfolio |
| Open Source Code | No | The paper does not contain any statement about making its source code publicly available, nor does it provide a link to a code repository. |
| Open Datasets | Yes | We select four diverse datasets from two categories to fairly appraise the new strategy. The Fama and French datasets as the first category have been recognized as high-quality and standard protocols in portfolio research (Fama and French 1992). FF25 includes monthly returns of 25 portfolios... and FF48 contains monthly returns of 48 portfolios... The second category comprises of the ETF139 and EQ181 datasets downloaded from the stock market. |
| Dataset Splits | Yes | In addition, for each strategy in our calculation we apply the rolling-horizon setting for the sequential out-of-sample performance evaluation (De Miguel, Garlappi, and Uppal 2009). In particular, from time tτ to tm, at each rebalancing time tk for k = τ, . . . , m, we first calculate the portfolio weight ˆωk based on the return data {Rl}k l=k τ+1. Then, we compute the realized out-of-sample net return ˆrk for the subsequent trading period. ... We use short sliding widows τ = 30, 60, 150 and 200 for the four datasets, respectively. |
| Hardware Specification | No | The paper does not provide any specific details about the hardware (e.g., CPU, GPU models, memory) used for running the experiments. |
| Software Dependencies | No | The paper does not list specific software dependencies or their version numbers used in the implementation or for running experiments. |
| Experiment Setup | Yes | We use short sliding widows τ = 30, 60, 150 and 200 for the four datasets, respectively. EGO, SS, RE and GO implement the growth optimal portfolio with the return target as the average of minimum and maximum estimated returns at each step. The relevant parameters for EGO, SS and RE are from n1 = 100, n2 = 5τ, n3 = 100 and n4 = n0.7. The risk-aversion coefficient in TZT, KZT, TZF and KWZ is three as the original papers. |