應用蟻群最佳化系統於臺灣股票市場投資績效之研究-以高薪100指數成份股為例

Abstract

本研究主要目的是應用蟻群最佳化來設計與建構一的投資決策系統，模型中以2013年臺灣高薪100指數成份股作為研究資料，實驗期間為民國103年1月2日至民國104年8月29日，共計20個月的數據資料，利用浮動視窗的觀念進行學習與應用，以五個月為一個學習期進行系統測試，並以應用期作為佐證。本研究設計三種投資決策模型，模型中以平滑異同移動平均線MACD、相對強弱指標RSI及隨機指標KD等技術分析作為擇時能見度，以本益本及市價對帳面價值比基本分析作為選股能見度，運用選股與擇時決策買賣策略選擇累積費洛蒙氣息，根據ACO之路徑選擇進行投資。研究結果顯示，三種投資決策模型，以同時考量擇時及選股之模型，在研究期間之投資報酬高達21.8％，其獲利表現優於大盤的投資績效，且遠超過銀行定存之獲利。 ACO系統參數值測試實驗結果顯示費洛蒙揮發系數設為0.5,0.6,07時，ACO系統搜尋最佳投資組合的能力為最好，且費洛蒙刺激係數α、與能見度係數β、γ值的變動會影響系統投資決策績效，當α=1、β=1、γ=1，為考慮選股與擇時的投資決策模型系統投資績效最高，α=1、β=1、γ=0時，為不考慮擇時的投資決策模型其投資績效最差。

This research is to study the optimization of Ant Colony Optimization(ACO) system parameters aiming at constructing a rational investment decision-making system for Taiwan stock market, taking Taiwan High Compensation 100 Index (Taiwan HC 100 Index) as the chosen sample. The experiment term is adding up to the data for 20 months from January, 2013 to August, 2014. During the experiment term, 5-month moving time window is chosen to train and test the data and system. The presented model takes into account the technical indicators which include MACD, RSI and KD line to represent visibility of timing, and PE ratio and BB ratio to represent the visibility of individual stock. With the visibility of timing, individual stock and phenomenon, the optimized ACO path is calculated and identified. The result shows the profit gaining performance of this system is 21.8% in the study period which is prior to TAIEX and the time deposit. With the Fellow Mongolia evaporation rate, 0.5, 0.6 and 0.7, the investment achievements of ACO are in the lead. Alpha(α), Beta(β) and Gama(γ) the values change will affect the investment of achievements performance of the ACO system accordingly. When α=1、β=1、γ=1，the investment achievements of ACO are highest taking into account timing and individual stock; while α=1、β=1、γ=0，the investment achievements of ACO the worst taking into account individual stock only.