A Hybrid Optimization Framework with Dynamic Transition Scheme for Large-Scale Portfolio Management

Round 1

Reviewer 1 Report

Li and Tam propose in the manuscript a new hybridized algorithms and demonstrate its efficiency on a range of benchmarks, both generic and from the financial optimization domain. The manuscript captures the challenges and technical background appropriately, maybe in too fine detail. The benchmark problem represent important challenges and their proposed algorithm performs well.

Clarification and changes are needed for:

* “Yet when hybridizing different algorithms, it should not unconditionally exchange all information between meta-heuristic algorithms for which the diversification of population is reduced.” This sentence is quite unclear (yet central to the paper). One can somewhat deduce its meaning from later portions of the paper but clarification in place would be good.

* the financial data set used is described as “from Yahoo Finance” but since modifications have taken place by the authors, providing the actual data set for download is important.

* the authors compare their proposed algorithms to a range of other algorithms representing other state of the art optimization techniques. However, it is not clear if the implementations chosen are indeed state of the art. Quite a few of them are also performance sensitive to configuration/settings. More transparency is requested on this point.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

1. The paper length is still wrong. Please consider to reduce the length.

2. A major concern of this paper is the encoding of solution. Please illustrate the solution in a figure.

3. A significant contribution of this paper might be the dynamic contribution algorithm. Hence, it is worthwhile to give a numerical example of it.

4. Because the proposed framework is the combination of different algorithms, 

5. It is good to include the CEC 2019 test cases in this paper due to there are some consideration of these dataset. However, does this portfolio selection problem is also the continuous problem?  

6. It is noticeable that when the number of stocks is increased, the performance of other metaheuristics is decreased while the proposed even perform better. Please discuss the reason.

7. The computation time is not included into the tables. Please put the CPU time on the compared algorithms.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf