Time-series gradient boosting tree for stock price prediction

Kei Nakagawa, Kenichi Yoshida

Abstract: We propose a time-series gradient boosting tree for a dataset with time-series and cross-sectional attributes. Our time-series gradient boosting tree has weak learners with time-series and cross-sectional attributes in its internal node, and split examples based on similarity between a pair of time-series or impurity between cross-sectional attributes. Dissimilarity between a pair of time-series is defined by the dynamic time warping method. In other words, the decision tree is constructed based on the shape that the time-series is similar or not similar to its past shape. We conducted an empirical analysis using major world indices and confirmed that our time-series gradient boosting tree is superior to prior research methods in terms of both profitability and accuracy.