Comparison between multiple linear regression and multi-layer perceptron neural network for predicting stocks from mining sector in LQ45

The development of mathematics means that there are more choice of techniques to predicting stock price movements. There are many tools for prediction in mathematics but on this study we adopt Multiple Linear Regression model and Multi-Layer Perceptron Neural Network with three variables input and one variable target. We use open price, high price, low price, and close price of mining sector stocks data in LQ45 from February to July 2020. Multiple linear regression is a further development of simple linear regression, the differences of this method is that multiple linear regression has more than one independent variable. Multi-layer perceptron is a feedforward neural network. Generally Multi-layer perceptron has three layers; an input layer, hidden layers and output layer, and in this study using 3 hidden layers. The results of this study are the multiple linear regression for LQ45 produces RMSE of 0.0161656 and multi-layer perceptron neural network for LQ45 produces RMSE of 0.0868014.