本研究採用台灣大盤股價指數及成交值所轉換的18個價量技術指標做為輸入參數,以期末資金最大化做為適應度函數,應用遺傳演算法(Genetic Algotithms)建構買賣決策類神經網路(Neural Networks)。研究結果顯示,本研究所比較之四種交易策略:遺傳神經網路策略(Genetic Neural Networks, GNN)、遺傳邏輯規則策略(Genetic Logic Rule, GLR)、單一遺傳邏輯規則策略(Single Genetic Logic Rule)及買入持有策略,在測試範例期間的平均年獲利率分別是10.27%、2.02%、-0.05%與-7.2%。另外由GNN、GLR及SGLR的風險評估得知,其高於買入持有策略的機率分別是91.77%、80.51%及79.39%。因此,GNN是一個穩定且有效的台灣股市交易策略。
This paper used 18 kinds of price and volume technical indices transferred from the Taiwan stock price index as the input parameters, the maximization of the final capital as the fitness function, the genetic algorithms as the optimization tool to construct the trading system based on neural networks. The results showed that the four kinds of trading strategies, the Genetic Neural Networks strategy(GNN), the Genetic Logic Rule strategy(GLR), the Single Genetic Logic Rule(SGLR), and the buy and hold strategy, in test period produced the average year profit rate respectively are 10.27%, 2.02%, -0.05%, and -7.2. Moreover, compared the risk assessment of the GNN, GLR and SGLR strategies, the probability of average year profit rate higher than the buy and hold strategy rate respectively are 91.77%, 90.51%, and 79.39. Therefore, the GNN trading system is a not only effective but also stable Taiwan stock market of trading system.
為了持續優化網站功能與使用者體驗,本網站將Cookies分析技術用於網站營運、分析和個人化服務之目的。
若您繼續瀏覽本網站,即表示您同意本網站使用Cookies。