Index Futures Trading With Stable Profits Using Deep Learning Models

Abstract

Investors can obtain high returns with a small amount of money through margin trading in the futures market but they also carry the risk of significant losses due to highly leveraged positions and large contract sizes. To mitigate these risks, investors often rely on stock market forecasting tools, and deep learning has become increasingly popular among researchers as an effective method for predicting market trends. This research presents a transfer learning approach for deep learning models to predict monthly average index of Standard and Poor’s 500(S&P 500) and Taiwan Stock Exchange Capitalization Weighted Stock Index(TAIEX) and use it to simulate trading E-mini S&P 500 and Mini-TAIEX futures contracts for evaluation. It conducts three experiments to show that the approach can gain stable profits. The first experiment is to analyze the results of different types of data preprocessing and trading strategies and find a general one for the following experiments. Second, we compared the results between the original and transfer learning methods to prove that our techniques are able to get consistent earnings. Finally, we proposed some ensemble models and found that the ensemble methods were more effective and stable to make profits.

Alternate abstract:

投資者可以透過期貨市場的保證金交易以小額資金獲得高回報，但也因高度槓桿化 的部位和大型合約的大小而承擔著重大虧損的風險。為了減輕這些風險，投資者通常 依靠股市預測工具，而深度學習已成為研究人員越來越受歡迎的預測市場趨勢的有效 方法。本研究提出了一種轉移學習方法，用於深度學習模型預測標普 500 指數和台灣 加權指數的月均值並用於模擬交易小標普和小台指期貨合約進行評估。它進行了三個 實驗，以顯示該方法可以獲得穩定的利潤。第一個實驗是分析不同類型的數據預處理 和交易策略的結果，並找到一般性的策略用於以下的實驗。第二個實驗是比較原始和 轉移學習方法之間的結果，證明我們的技術能夠獲得一致的收益。最後，我們提出了 一些集成模型，並發現集成方法對獲利更有效且更穩定。