Nowcasting vs Forecasting for Financial Markets

When I was doing the first project, one topic caught my attention: Nowcasting for financial markets.

  • Traditionally, quant strategies have focused on forecasting prices, using structured data to make long-range predictions. However, these forecasting models do not adjust quickly to changing market conditions. This disadvantage becomes particularly serious during the epidemics.
  • In contrast, nowcasting models use unstructured datasets to make direct measurements and short-range predictions, which are more reliable and make full use of millions of recent observations.
    Then I have questions like what machine learning models are suitable for making nowcasting? Or what changes can we make to the forecasting model to realize nowcasting?

Thank you that is another good question, I will once more give five participation points to anyone who answers this question.

Changing market conditions do imply that there is any uncertainty in the events and may affect the prediction by a great margin, which as you mentioned, is like epidemics.
To account for such uncertainty, one could try thinking of these as required outliers that we need to check and be confident in dealing with them properly. Using this idea, models that are good with outliers can be preferred.
Another idea is to use Fuzzy Systems, which does provide an advantage in modeling fuzzy information and combining them with Machine Learning or Deep Learning models. One such model is Adaptive Neuro-Fuzzy Inference System (ANFIS).
I am looking forward to this discussion here. It’s very interesting!