PREDICTION OF STOCK PRICES VIA RECURRENT NEURAL NETWORKS WITH LSTM (LONG SHORT-TERM MEMORY) ARCHITECTURE
Hyacinthe Kouassi Konan, Francis Adles Kouassi, Mamadou Tidiane Coulibaly, Olivier Asseu · Far East Journal of Mathematical Sciences (FJMS) · 2020
Recurrent Neural Networks (RNNs) have gained popularity in the field of prediction of stock market prices due to their extraordinary performance in time-sequential tasks. Today, LSTMs are one of the most popular sequential deep learning neural network architectures. The main idea behind LSTM neural networks is to allow the network to “forget” or ignore certain past observations so that it can give weight to important information in the current prediction. In this article, we propose to use LSTMs for predicting the stock values of Facebook, Tesla, Apple and Microsoft, quotes that are widely followed by investors around the world.