The Application Of Artificial Bee Colony Algorithm In Investment Portfolio Problem

The modern portfolio theory originated from Markowitz' mean-variance model.Many scholars have developed different ways to improve and solve the model since the mean-variance model was proposed.The purpose of the establishment of the investment portfolio model is to maximize the investment income and minimize the risk.Now,there are two main research directions: one is to optimize the model itself and the other is to use more effective method to solve the problem.This paper proposes a more practical mean-variance model,in which the stocks are divided into different fields,and which is called the structured-MV model.We also improve the ABC algorithm,the improved algorithm called ABC-GL.The experimental result shows that the convergence speed of ABC-GL algorithm is better than the ABC algorithm.This paper also proposes the structured Mean-Absolute Deviation model,and then we prove the structured Mean-Absolute Deviation model has no different with the Mean-Absolute Deviation model.Finally,based on one hundred day's history price of 71 selected stocks,we use the ABC and the ABC-GL algorithm to solve the Markowitz' mean-variance model,the structured-MV model and the Mean-Absolute Deviation model.After that,we do the comparison between the solutions obtained by both algorithms on these models.We can achieve two conclusions based on the empirical analysis: ABC-GL algorithm is better than the ABC algorithm;the solution of the structured--MV model is better than the Markowitz' mean-variance model no matter use ABC or ABC-GL algorithm.