For quantitative investment, besides market information, it is also very important to collect other fundamental information, and the corresponding time series should be integrated into the forecasting model. What is the successful model? The key point is how many different information sources it integrates, not how advanced mathematical theory it uses. Taking simple linear regression as an example, if the prediction effect of the model is good, all parameters need to have strong prediction ability and low correlation; On the other hand, if the selected parameters are meaningless, then the model is useless even if it is applied to complex deep learning theory. Some companies in the United States not only use text information such as news to model, but also use port container images taken by Google satellite to model. What is the trend of commodity prices? Through the prediction of the number of commodity containers, good prediction results have been achieved.
Solving the model is actually as important as modeling. For example, there are many models in physics that can accurately describe reality, but they are still difficult to solve because of the lack of efficient scientific calculation methods, and so is quantitative trading. With the calculation, screening, optimization, back test and so on of parameters with huge calculation, how to solve it skillfully is a very profound knowledge. Simmons revealed that this famous renaissance company has a clear division of labor-physicists analyze data to build models, mathematicians build optimization algorithms and solve models, and computer programmers collect data from various sources.