Method and device for constructing deviation rectification tree model and method and device for evaluating intervention effect by using deviation rectification tree model

The embodiment of the invention provides a method and a device for constructing a correction tree model and evaluating an intervention effect by using the model. The method for constructing the tree model comprises the steps that a sample set is obtained, the sample set comprises index values of user behaviors of all users in two groups of users in m time periods, and the m time periods comprise m-1 preorder time periods and a target time period; one of the two groups of users is subjected to an intervention behavior during a target period of time. A regression tree is constructed based on splitting loss of node splitting, and the splitting loss comprises confusion entropies of two child nodes obtained through splitting; the confusion entropy of any node is the sum of the cross loss of each user falling into the node in the preorder time period, and the cross loss represents the difference degree between the historical index value of the user in the preorder time period and the mean value of the index values