Since using neural network can solve the problem of classification, the SVM and the decision tree these algorithms still have what meaning?

CodePudding user response:

Each model has its own assumption, each data set has its own distribution, the distribution of the assumption of the model with data sets fit, the higher the model performance, the better, the so-called "there is no best model, only the most appropriate model", such as a set of data interface is linear, and the number of small, in this case, you use the neural network is not worth the cost, on the contrary, with a simple LR will bring out the best in each other, hot, in recent years neural network for beginners a sham, feel no neural network to solve the problem, some even want to use neural network prediction of lottery tickets, but neural network are assumption, if the data is not in conformity with the corresponding assumption, neural network can do,