The Research Of Online Portfolio Selection Based On The Online Learning Algorithm

Nowadays online learning has been extensively studied in artificial intelligence and machine learning communities.In recent years,online portfolio selection based on online learning has become one of the most popular researches in the field of machine learning and artificial intelligence.Most existing online portfolio selection strategies,such as universal portfolio(UP)based strategies and EG based strategies always ignore transaction cost,and thus when they apply to the market,the total transaction cost increases rapidly and the final cumulative wealth degrades severely.To overcome the limitation,in this paper we will propose two kinds of new strategies which can apply in the online portfolio selection with transaction costs.First,according to the problems of high transaction frequency and high transaction costs in Cover's UP,we propose a new Semi universal portfolio strategy based on competitive solutions.Compared to UP,which will trade in every period,SUP will not trade when the net wealth gain in this period is less then the cost for the transaction.We show that the proposed SUP strategy is universal and has an upper bound on the regret.We present an efficient implementation of the strategy based on non-uniform random walks and online factor graph algorithms.Empirical simulation on real historical markets show that SUP can achieve significantly better performance.Second,we propose a ONS algorithm with transaction costs(ONS-Cost)according to Newton Steps.The ONS-Cost use the second order information and the loss term of transaction costs.The algorithms can control the trading volume and reduce the transaction costs too.Meanwhile,we present an efficient implementation of the strategy based on Newton step.Empirical simulation on real historical markets show that ONS-Cost can perform better than ONS,especially when the transaction costs exist.Research of this thesis has some theoretical significance to online portfolio selection research.It also plays an important role in guiding the actual portfolio selection in the financial industry.