Investment portfolio balancing: application of a generic self‐organizing fuzzy neural network (GenSoFNN)

Abstract

In contrast to short‐term stock trading, portfolio managers are interested in the medium‐ to long‐term peaks and troughs of the stock price cycles as signals to balance their stock portfolios – the predicted trough is the signal to buy the stock and the predicted peak is the signal to sell the stock. As statistical models are generally inadequate or incapable of providing such portfolio balancing signals, we propose using the generic self‐organizing fuzzy neural network (GenSoFNN)—a fuzzy neural system – as a tool for portfolio balancing. The network adopts the supervised learning approach to detect inflection points in the stock price cycles, and a modified locally weighted regression algorithm is employed to smooth the stock cycles. The GenSoFNN‐based portfolio balancing system was evaluated with experiments conducted using 23 stocks from the New York Stock Exchange and NASDAQ, and the results showed an average profit return of 65.66%. The contributions of the proposed GenSoFNN intelligent portfolio balancing system are twofold: it can be used as an efficient trading solution and it can provide decision support in trading via its generated rules. Copyright © 2009 John Wiley & Sons, Ltd.

A comparative survey of artificial intelligence applications in finance: artificial neural networks, expert system and hybrid intelligent systems

“…They finally claimed that the optimal investment proportions can thus be determined, according to different confidence levels. In the same domain, [199] proposed a fuzzy neural system for portfolio balancing using the generic self-organizing fuzzy NN (GenSoFNN). In their model, they used supervised learning approach in the network in order to detect inflection points in the stock price cycles.…”

Stock Portfolio Optimization Using a Combined Approach of Multi Objective Grey Wolf Optimizer and Machine Learning Preselection Methods

“…In research on financial market forecasting of artificial intelligence algorithms including artificial neural networks [20][21][22][23], recursive neural networks [24], genetic algorithms with neural networks, artificial [25,26] and support vector neural networks [27] have been used. Quek et al [28] proposed the use of fuzzy neural network (GenSoFNN) as a tool for stock portfolio balancing, which uses a supervised learning method to detect milestones in the stock price cycle and a modified weighted regression algorithm used to smooth the stock cycle. Hsu et al [29] used Markov chains and fuzzy theory to create a stock market index forecasting model.…”

Neuro-fuzzy time-series analysis of large-volume data

“…Hsieh et al () use wavelet transforms, a recurrent neural network and an algorithm based on an artificial bee colony (Karaboga and Basturk, ). Ballini et al (, ) and Quek et al () use fuzzy neuro systems to investigate risk analysis.…”