preview

When Popularity Of Machine Learning Models Increased, A Number Of Automated Trading Systems

Decent Essays
Open Document

When popularity of machine learning models increased, a number of automated trading systems were build around these models. But rst, let 's take a look at the history of machine learning models in the eld of nancial predictions. At rst, White (1988) applied arti cial neural networks
(ANN) to reveal nonlinear regularities in the IBM stock price movements. Subsequently, Kamijo and Tanigawa (1990) used a recurrent neural network for the recognition of price patterns in the
Japanese market. Cheng, Wagner, and Lin (1996) used an ANN to predict the weekly price direction of the 30-year U.S. treasury bonds and averaged an annualized return on investment of
17:3%. Later, A.-S. Chen, Leung, and Daouk (2003) predicted the direction of the return on the
Taiwan Stock Exchange Index and showed that the ANN based strategies outperformed the random walk model and the generalized methods of moments with Kalman lter. Despite the reported success, many researchers have shown that neural networks have some limitations. For instance,
Schoneburg (1990) pointed out that the performance of a neural network is very sensitive to its design. Lee, Oh, and Kim (1991) highlighted the slow convergence of the backpropagation learning algorithm (common method to train a neural network) and its convergence to local minimum due to the non-linear dimensionality of market data. Nevertheless, neural networks and its various extensions are still widely used.
Probably one of the most popular methods for

Get Access
Related
Get Access