The Trend is Your Friend: A Note on An Ensemble Learning Approach to Finding It

Abstract

The essential goal of trend-following investing is to precisely identify where the uptrend and downtrend are located. This paper thus provides a two-layer stacking technique, which is a novel ensemble learning approach, to predict such trends for the Taiwan Top 50 ETF. The proposed stacking technique stacks the predictors of support vector machine (SVM), multi-layer perception (MLP), adaptive boosting (Adaboost), and extreme gradient boosting (Xgboost), presenting empirical results whereby following the trends obtained from the stacking technique can generate positive returns and beat both conventional moving-average crossover and buy-and-hold strategies.