Paper Title
Trend Prediction Method with Feature Selection based on SVM and Ensemble Learning

Abstract
In this paper, we proposed a trend production model with feature selection method which can improve the generalization performance and reduce calculation time while kept the accuracy. The proposed method is a hybrid model combine the filter and wrapper which can improve the calculate efficiency. The experimental results show that the prediction accuracy of using features is better than that based on historical data alone. The accuracy of the proposed method is 0.9% and 11% higher than the conventional best model GA-SVM in terms of f1-score and prediction error. The generalization performance of the proposed method has a 33.1% improvement in cross-validation variance compared to GA-SVM, and the computation time is saved by 71%. Keywords - Trend prediction, Feature selection, Random forest, Ensemble learning, Generalization performance, Genetic algorithm.

Author - Chen, Hongbin, Murata, Tomohiro

Published : Volume-5,Issue-12  ( Dec, 2018 )

	|		PDF	|	Viewed - 46	|	Published on 2019-02-21