Incorporating I Ching Knowledge Into Prediction Task via Data Mining

Introduction

I Ching is an ancient Chinese document which contains a wide variety of rules. I Ching divination uses Ying and Yang and the Four Signs to explain the laws of the changes in the world and predict all kinds of affairs. Nowadays, researchers leverage the I Ching numerical hexagram model to stock market prediction (Guo & Lu, 2020), and apply hexagrams to business guidance (Chen, 2021). However, previous I Ching-based stock prediction mechanisms have several weak points: Small scale stock market data lead to unconvincing statistical results, oversimplified indicators lead to incomplete initial feature, and many abstractive I Ching concepts cannot transform into available modus.

To solve the above problems, the authors integrated I Ching prediction mechanism and machine learning, presenting a prediction model based on I Ching. Traditional I Ching predicting methods mainly utilize the explanations of Yao and hexagram to guide people make reasonable judgments.

Feature selection is an important step in building the I Ching prediction model. The selection algorithm can filter out the features with higher importance from the high-dimensional feature set. Currently, many feature selection methods are widely used in optics science (Huang & Liu, 2021) and medical science field (Liu et al., 2019). Classic integrated feature selection algorithms such as random forest (RF)(Breiman, 2001) integrate the results of feature selection methods by training multiple features. In real-world applications, the Three Vitals especially reflect the influence of nonhuman factors such as national policy and environmental conditions.

Figure 1.

Prediction model framework

Note. Left: The generator receives unprocessed feature sets. The authors leverage Three Vitals and Six Yao to fine-tuned RF algorithm and sliding windows technique, which improves the prediction capacity and flexibility. Right: The decoder receives the hexagram as input, and extracts hexagrams processer and explainer from “ZhouYi.”

The thought of Ten, Chi, and Jin in the “ZhouYi” is also called the Three Vitals, which represent heaven, earth, and human. The Three Vitals assert that human beings are a separate entity isolated from heaven and earth, but also recognize that human behavior can affect the operation of the entire world. Specifically, the authors propose a feature selection method based on machine learning to rank feature importance in the first place, then leverage the Three Vitals concept to filter out the most important feature in current scene. Secondly, the authors utilize sliding windows algorithm to transform features into hexagram; then, they determine the original hexagram and change the hexagram according to the I Ching hexagram generating method. Furthermore, the authors utilize the I Ching explaining method to determine stock trend and finally get the prediction result by comparing predict value and tag value. Combining I Ching and machine learning gives the model more flexibility. The experiment result shows promising outcomes in the stock market prediction task, compared to popular machine learning and deep learning algorithms such as support vector machine (SVM), eXterme Gradient Boosting (XGBoost), attention long short-term memory (LSTM), and gate recurrent unit (GRU).

Background

The concept of Yin and Yang originated from the Chinese I Ching culture system and it is the foundation of ancient Chinese philosophical thought (Li, 2014). Yin and Yang theory shows two different states and teaches people to see the world’s problems in two ways. In real-world applications, Yin and Yang can correspond to both good and bad aspects of one thing. Any complex thing can be restricted and divided by the Yin and Yang.