Investment portfolio optimization using local version particle swarm optimization with mutation
Due to development of high-power computers, heuristic algorithms are applied broader at present, especially in financial engineering. Particle Swarm Optimization, or PSO, is one of the popular heuristic algorithms, and it has been proposed with multiple forms of variants. In this article, we will pr...
Saved in:
| 主要作者: | |
|---|---|
| 其他作者: | |
| 格式: | Final Year Project |
| 語言: | English |
| 出版: |
2016
|
| 主題: | |
| 在線閱讀: | http://hdl.handle.net/10356/68051 |
| 標簽: |
添加標簽
沒有標簽, 成為第一個標記此記錄!
|
| 總結: | Due to development of high-power computers, heuristic algorithms are applied broader at present, especially in financial engineering. Particle Swarm Optimization, or PSO, is one of the popular heuristic algorithms, and it has been proposed with multiple forms of variants. In this article, we will present a new variant, called local version PSO with Random topology and mutation (RM-LPSO), to solve investment portfolio optimization (PO) problems. Markowitz constrained model will be set as our fitness function for PO. RM-LPSO uses local version PSO with random topology, and also particles can mutate during moving in searching region. Moreover, some adjustments based on characteristics (useful and useless assets) of PO problems are made, to simplify our computing and increase accuracy. We introduce another PSO approach Dynamic Random Population Topology with same degree (DRTWPSO-D) to compare to. From our experiments, RM-LPSO without such adjustment shows good result in dealing with stock market whose number of assets is high; however, it does not show priority in markets with fewer assets. Comparatively, RM-LPSO with adjustment performs well and it achieves better result in all five stock markets. Thus we can say, RM-LPSO is good enough to solve larger stock market PO problems, and adding the adjustment based on useful and useless assets will also improve RM-LPSO performance. |
|---|