Automatic Stock Market Prediction using Novel Long Short Term Memory Algorithm compared with Logistic Regression for improved F1 score

Abstract

This work provides a comparative study of improved F1 score in stock market values using a novel long short term memory algorithm (LSTM) which is compared to Logistic Regression algorithm. Materials and Methods: Novel Long Short Term Memory (<tex>$N=10$</tex>) and logistic regression algorithm (<tex>$N=10$</tex>) were iterated to improve F1 score for stock market predicted values. Two algorithms are simulated by varying NLSTM and logistic regression parameters to optimize pH. Sample size is calculated using Gpower 80&#x0025; for two groups and there are 20 samples used in this work. Results and Discussion: LSTM has notably better accuracy percentage (68.24&#x0025;) compared to logistic regression accuracy (53.71&#x0025;) with 0.407 (<tex>$p &#x003E; 0.05$</tex>). Conclusion: Long short term memory algorithms help in predicting automatic stock market prices to improve F1 score.