Machine learning strategies for stock market predictions

Predicting stock prices is difficult since they vary randomly. Stock purchasers can increase returns by reducing risk. This thesis presents a deep learning method for stock market prediction by suggesting the combination of deep learning with weight memory to better forecast stock prices. 1D-CNN with LSTM, GRU, or SimpleRNN are these methods. Error metrics have been studied to reduce predicting mistakes, which may cost the stock market millions. The best prediction model with the lowest error metric will answer future stock market trading with organization transactions. GRU and 1D-CNN have the lowest root mean squared error, giving them the best prediction method. SimpleRNN outperforms LSTM in the mean squared error. GRU and 1D-CNN also excelled in mean square error and meant absolute rate. The GRU combination won again with the highest R2 score. The findings of this study have been compared to other studies from the literature review and show that the proposed method performs better, meaning that GRU and 1D-CNN have the lowest error potential.