Treffer: A hybrid predictive prototype for portfolio selection using probability-based quadratic programming and ensemble artificial neural networks.

Often, investors are limited by cognitive and emotional biases in their decision making which leads to poor portfolio investment choices. Robo-advisors can assist in overcoming these biases. This paper seeks to develop a financial robo-advisor prototype based on hybrid programming. It uses ensemble artificial neural networks to predict portfolio returns and variances with input nodes of Ornstein-Uhlenbeck processes (OU) and geometric Brownian motion (GBM) processes' estimates. The results are subsequently channeled into a probability quadratic optimization algorithm which considers target return probability and value-at-risk constraints as proxies for investor's risk tolerance so as to provide the optimal portfolio allocation strategy that minimizes portfolio risk given a prescribed investment horizon and target return. The results show that the ensemble artificial neural network method implementation accurately predicted the level of 2 of 5 assets and the trends of the remaining 3 assets. However, it yielded low standard deviations and low returns compared to the OU and GBM estimates for short horizons. The quadratic optimization algorithm supported investment in shorter time horizons since portfolio risk was lowest. Diversified allocation was achieved in the shorter time horizons while longer horizon allocations were biased towards assets with lower standard deviations. The lowest risk portfolios were the ones with a lower certainty probability for the target return and vice versa. This paper clearly demonstrates that ensemble methods are accurate in prediction, and that a hybrid programming paradigm effectively leverages the strengths, speed and functionality of different programming languages -- an elixir for multifaceted dissociable programming problems. [ABSTRACT FROM AUTHOR]

Copyright of Orion is the property of Operations Research Society of South Africa and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)