Developing a stock price predictive model

As stated earlier, the stock market price has severe volatility and historical perspective, which make it suited for time analysis. This also challenges those classical algorithms, since long-term dependencies cannot be availed using those algorithms.

As outlined in following diagram, first we collect historical financial data. The data is then converted into a time series after the necessary preprocessing and feature engineering. The resultant time series data is then fed into the LSTM to carry out the training. The following diagram illustrates this:

High-level data pipeline of the prototype used for this project

Therefore, we will be using LSTM not only because it outperforms classical algorithms but also because we can solve long-term dependencies with it. Consequently, our project will have the following steps:

1. Load and preprocess the data, and split it into train-and-test sets
2. Train the LSTM model with the data
3. Evaluate the model on test data
4. Visualize...