Stock Trading Strategy And Adaptive Rules Discovery

Since Fama and French put forward the efficient markets hypothesis. Thediscussion on stock markets efficient become a focal point in recent researches. In thisresearch, the efficiency of Chinese Stock Exchange verified from two aspects, such as:fundamental analysis and technical analysis. Finally we come into the conclusion thatour research deny the efficient markets hypothesis. Especially, on comparison withthe former researches, the adaptive trading model provided in this paper can makeexcess profits in different market phases.In first part of this paper, the cross-section methods was used to do fundamentalanalysis，the results indicate that the book to equity, size and other account factorshave strong explanation power on expected stock return. The result shows specificfactors`explanation power on expected return in Chinese market is different withAmerican market. Furthermore, the reversal take place on different periods oncorrelation analysis based on the book to equity. On the other words, the concept drifttake place among different periods and regions, that is the “anomalies”.At the second part, in consideration of the importance of rule discovery andconcept drift in trading model construction. In this research, the eXtended ClassifierSystems (XCS) was applied to develop an intelligence trading model--eTrend. Thatbased on the current stock market condition and the trend following trading rules tomake buy/sell decisions in test period. During the trading process, eTrend adapt to theenvironment quickly and save the experience for making optimal decision. ThreeIndex in Shanghai Exchange were selected as the testing sample. In the12-yearstesting periods, the result shows that, no matter in Bull or Bear phase, the model canachieved high excess return without the transaction. Moreover, the downside risk low,that Sortino rate is close to1.0. In addition, the experiment include a comparisonbetween Decision Tree (DT), Artificial Neural Network (ANN) algorithms andeTrend on effectiveness, the result is obviously that the eTrend has more advantages.The mainly contribution of this paper is twofold.(1) From the academic aspect,the fundamental analysis result is interesting, that the concept drift phenomenon existsin financial market. That provide an evidence for the importance of the concept drift;eTrend is one of the best solution which accomplish concept drift and rule discoveryboth.(2) From the investment practice aspect, this paper provide a new direction of investment strategy construction which helpful for the investors. Which combiningthe investment rule in practice and artificial intelligence for constraint learning thatlead to impressive result. That a good sample for investors.