Predicting stock’s price direction is a
challenging task due to its noisy and non-stationary nature.
The aim of this study is to create a short-term stock prediction
model. The model was built in two stages. In the first stage,
this study implements three different dimensionality
reduction methods (Lasso, Elastic Net PCA) to reduce the
number of original features In second stage, the informationrich feature set which resulted from the first stage are used as
input into an LSTM to generate the Trading signal. The result
shows that the combination of PCA and LSTM outperform
the other two combinations.The combination of PCA-LSTM
reached 71?curacy while Lasso and ElasticNet accuracy
consecutively 60% and 55.7%. Moreover, in this study PCALSTM model only use one principal component with 32%
variance explained to give such accuracy, while the other two
models use more than half of the original features. The overall
result suggests that for predicting stock price’s movement,
feature extraction is a better option than feature selection for
preprocessing step
Level
Internasional
Status
Dokumen Karya
No
Judul
Tipe Dokumen
Aksi
1
ICEECS-RUVITA-PRN.pdf
[PAK] Full Dokumen
2
cover-vol1-september-2019.pdf
3
Predicting Stock’s Movement Using Unidirectional LSTM and Feature Reduction The Case of Indonesia Stock.pdf
