Research on Factor Interaction Effects and Nonlinear Relationships in Quantitative Models

In the process of constructing quantitative models, the majority of existing fundamental quantitative stock selection models typically assume a linear relationship between factors. This simplifies the construction and interpretation of the model. However, it also ignores the complex interaction effects and nonlinear relationships that may exist between factors. Such relationships may prove to be of significant consequence with regard to the predictive capacity and accuracy of the model. In this work, the gradient boosting machine (GBM) is employed to investigate the nonlinear relationships and interactions between factors, with a view to constructing more complex quantitative models of fundamentals. Gradient boosting machines incrementally enhance the precision of the overall model by iteratively assembling a sequence of predictive models. It is capable of capturing the intricate nonlinear relationships between factors, as well as the interactions between factors, which are challenging to discern with traditional linear models. The application of GBM in a quantitative model enables a more precise representation of the interactions between financial market factors. By progressively optimising the loss function, GBM can account for the influence of diverse factor combinations, thereby enhancing the model's predictive capacity.