Through testing all commonly used model architectures (CNN, LSTM, or ARIMA Model) in the field on randomly selected periods of time on past global stock markets such as NASDAQ, S&P 500, How can ML model architecture/structure combination be best optimized for stock market profit and safety (based on standard regression model test measures).