Study on parametric model and nonparametric model

In the era of big data, we often face hundreds of millions of data, accompanied by high-dimensional variables. Nowadays, many academic and technical fields are committed to solving the problem of model construction for big data, such as neural network and deep learning. However, for the financial and business fields, how variables affect the response variables (interpretability), whether the model is reliable and other factors are particularly important. Especially when our data set contains only a few variables, the selection of characteristic variables should be more rigorous, and the quantification of the relationship between variables and response variables is also an important indicator, At this time, we can use more sophisticated model exploration methods. Generalized linear model and generalized additive model are respectively extended from linear model and additive model, which can be more widely applied to data with different distributions, and the model is gradually optimized with the help of likelihood ratio test. In this paper, from the perspective of parametric model and nonparametric model, taking generalized linear model and generalized additive model as examples, with corresponding cases, the model exploration and optimization methods are briefly introduced.

Experience it now and start the journey of digital transformation !