Online Learning with Radial Basis Function Networks

TL;DR

This paper introduces an online learning approach using Radial Basis Function Networks for financial time series forecasting, effectively handling nonstationarity and concept drift to outperform traditional methods.

Contribution

It presents a novel online RBF network method that combines feature transfer and sequential optimization for improved multi-horizon financial forecasting.

Findings

Outperforms random-walk baseline in forecasting accuracy

Better handles nonstationarity and concept drift in financial data

Demonstrates superior performance over batch learners

Abstract

Financial time series are characterised by their nonstationarity and autocorrelation. Even if these time series are differenced, technically ensuring their stationarity, they experience regular covariate shifts and concept drifts. Against this backdrop, we combine feature representation transfer with sequential optimisation to provide multi-horizon returns forecasts. Our online learning rbfnet outperforms a random-walk baseline and several powerful batch learners. The rbfnets we formulate are naturally designed to measure the similarity between test samples and continuously updated prototypes that capture the characteristics of the feature space.