Stock Prediction on Historical Data based on SGD and LSTM

Stock market prediction is not an easy job this Stock market prediction is not an easy job, it requires a lot of extensive and intensive calculations, analysis, human and computer interaction. Now, we know the nature of stock prices, they change almost every day and almost every moment. The usual conventional methods that we use for batch processing cannot be utilized efficiently for stock market prediction. So in this paper we are proposing a machine learning algorithm that basically utilizes some kind of neural networks, usually Recurrent NN (RNN) known as Long Short Term Memory (LSTM), where the weights and biases are adjusted as per the singleton data points using something called SGD. This will contribute in providing more specific results when compared to already existing algorithms for stock-price prediction. We will then obtain a network which would be trained and evaluated for even higher accuracy with different variant of data sizes, and the results would then be analyzed and performed against an Artificial Neural Network.