| Summary: | This paper explored the process of investment management in both theory and practice in China's mutual fund industry and reviewed the applications of artificial intelligence including Rule-based Expert Systems, Genetic Algorithms, Artificial Neural Network, and Support Vector Machines in financial forecasting, asset allocation and stocks selection.
This study proposed the use of artificial neural network for stock selection which classifies stocks into undervalued stocks (+1), neutral stocks (0) and overvalued stocks(-1) in China's market. The neural network used in this study is a multiple-layer feedforward neural network which uses a Levenberg-Marquardt accelerated training algorithm. There were three groups of input variables in this study. One unprecedented input was proposed in this study name analysts recommendations.
This study found that the traditionally managed equity fund average performance did not beat the market. It is also conclude that the artificial neural network in stock selection can be used to improve the current investment management process. This study also found that the use of analysts recommendations as input variable to the neural network was proved as ineffective to improve the stock selection performance; and, a misclassification problem due to too many input variables.
|
|---|