METHOD OF PREDICTING STOCK INDEX BASED ON NEWS ARTICLE ANALYSIS BY USING ARTIFICIAL NEURAL NETWORK MODEL AND APPARATUS THEREOF

摘要

The present invention relates to a method of predicting a stock index based on news article analysis by using an artificial neural network model and an apparatus thereof and, more specifically, to a method of predicting a daily stock index from a machine-learnt artificial neural network model by inputting a plurality of new articles into the model, and an apparatus performing the method. According to the present invention, from a news article, instead of extracting a relation tuple (O1; P; O2) having a structure similar to a subject (S), a verb (V) and an object (O) like in Open IE, word level information of which the level is lower than the relation tuple is extracted, and thus, input with a reduced information loss is provided to an artificial neural network. Since a news article title is data which is relatively short and has a low used word count deviation, work level information is used as input instead of applying the form of a relation tuple, and thus, an information loss can be significantly reduced and a beneficial result can be created. Moreover, since a recurrent neural network (RNN) or a long short-term memory (LSTM) model which is a kind of RNN is used as an artificial neural network, sequentially appearing data such as words of a news article can be more accurately processed, and thus, a more reliable result about a stock index can be derived.;COPYRIGHT KIPO 2020