Study on the Variation of the A Fund of the Pension System in Chile Applying Artificial Neural Networks

摘要

In Chile, all people have a legal obligation to enter and contribute to a pension system, capitalizing their savings through profits from investment funds at their reference value. The profits obtained depend on the prices of the financial instruments of the investment firms that manage these funds, and determine the amount contributors will receive at the time they collect their pensions, as well as the positive or negative variations in the value of their shares. The objective of this study is to evaluate the predictive capacity of the artificial red neural Red Ward Model by evaluating the percentage of prediction of signs and the weekly profitability the value share of Fund A of the Chilean pension system. In this research, we apply models based on neural networks to the Chilean pension system, as well as predictions of the weekly variations in the fund, to obtain an improved growth forecast. Our research shows that in five of the eight months we considered, a percentage higher than 65% was obtained, all the models had statistical significance, and most active investment strategies are superior to passive ones. Finally, we conclude that the built network has a strong predictive capacity, and that the use of artificial neural networks for the prediction of variations in financial values is a viable alternative since the results obtained are consistent with other existing methods such as the Vector Support Machine approaches.