A Computing Method for Solving Mean-absolute Deviation Portfolio Selection Model

In this paper, the model is reduced and solved by a pivoting-based algorithm proposed by the authors. This algorithm is simpler than the widely used simplex method and requires less amounts of computaion. The experimental results by using 70 weekly data of 1 072 stocks from Shanghai and Shenzhen stock markets of China indicate that 1 274 pivoting operations are required to obtain 20 different optimal portfolios if no upper bound restrictions for every stocks and 1 570 pivoting operations are required if the ratio of each stock is no more than 10%, in which one pivoting operation required about 1 141×71 andditions and multiplications.