Online Investment Strategy Based On Trend And Reversal

Online portfolio selection is a fundamental issue in computation finance.Experts in the field of artificial intelligence and machine learning has paid a lot of attention to this issue.Based on Capital Growth Theory and online learning theory,online portfolio selection usually designs reasonable online learning algorithms to guide the construction of portfolios.According to the existing research,there exist both trend(or momentum,inertia)effect and reversal effect of stocks prices.Some scholars have separately constructed the follow-the winner strategies and follow-the loser strategies,which had achieved good performance.Although existing strategies performed not bad on some real data sets,only one effect was used to construct the online portfolio selection models,which may not correspond to the reality that the market environment may exist both trend and reversal effects.In order to simulate the market,this article proposed Online Trend without Reverse(OLR)and Online Reverse without Trend(OLT).They can grasp the market changes and capture the trading opportunities.Take the trend effect and the reversal effect into account,these two algorithms were based on the PA(Passive Aggressive)algorithm.By using the convex optimization technique,this article obtained two online portfolio selection models and tested the models on the real data.From the empirical results,the three strategies performed better than most strategies tested,especially Online Trend without Reverse(OLR).In addition,the parameter sensitivity test was also carried out on the parameters involved in the algorithms,which indicated that the algorithms were stable where the parameters changed.