Selecting Data Adaptive Learner from Multiple Deep Learners using Bayesian Networks

TL;DR

This paper introduces a novel approach combining multiple deep learners and Bayesian networks to improve time-series prediction accuracy, especially applied to financial data like the Nikkei 225 index.

Contribution

It proposes a data-adaptive method that dynamically selects the most appropriate deep learner for time-series prediction using Bayesian networks and clustering.

Findings

Enhanced prediction robustness for financial time-series.

Effective selection of deep learners based on Bayesian posterior probabilities.

Demonstrated improved accuracy on Nikkei 225 index data.

Abstract

A method to predict time-series using multiple deep learners and a Bayesian network is proposed. In this study, the input explanatory variables are Bayesian network nodes that are associated with learners. Training data are divided using K-means clustering, and multiple deep learners are trained depending on the cluster. A Bayesian network is used to determine which deep learner is in charge of predicting a time-series. We determine a threshold value and select learners with a posterior probability equal to or greater than the threshold value, which could facilitate more robust prediction. The proposed method is applied to financial time-series data, and the predicted results for the Nikkei 225 index are demonstrated.