Comparison of Grid Search and Evolutionary Parameter Optimization with Neural Networks on JCI Stock Price Movements during the Covid 19

Abstract

This study aims to determine the effect of covid 19 on the movement of the JCI Stock Price by testing various combinations of the input variables of closed price stock data on the JCI. The analysis is carried out to find the best RMSE value from the combination of these input variables using the neural network method. The best RMSE results are compared using the optimization of grid search and evolutionary parameters. The data used in this study was taken from the Yahoo.finance.com page on the JCI Historical Data, during the covid pandemic, from 12/11/2019 to 12/30/2021. The data obtained are 509 records. The input variable used is the closing price data (closed price) as a target. The preprocessing data used are data cleansing, filtering, and windowing until seven days before. The results obtained an RMSE value of 0.104 five days before Close t (P=5), training cycle 9000. Momentum 0.9 and learning rate 0.2 is then optimized using the grid search parameter to produce RMSE 0.101, training cycle 100. Learning rate 1 and momentum 0.1 are then compared with evolutionary parameters, which make RMSE 0.103 at learning rate 0.029, momentum 0.68, and training cycle 86. Based on this research, optimizing grid search parameters produces better RMSE than evolutionary parameter optimization. This small RMSE result shows that investors are still safe to invest.