The invention discloses a method for constructing TV user portraits. The method comprises the following steps: step 1, collecting data of TV terminal users through a data platform, and analyzing and classifying the data; step 2, pre-defining a TV user portrait tag; step 3, classifying data of type B, and constructing a first level tag of the TV user portraits of the data of type B; step 4, classifying data of type C, and constructing the first level tag of the TV user portraits of the data of type C; step 5, constructing second level tags of the TV user portraits of the data of type B and type C; step 6, merging and counting each type of first level tags and each type of second level tags of the TV user portraits; step 7, analyzing program type preference attribute data of the TV users, and constructing TV user portrait tags; and step 8, updating the pre-defined TV user portrait tags into the constructed TV user portrait tags. According to the method disclosed by the invention, the viewing behavior data of the TV users are analyzed to master the preference of the users so as to provide better personalized service for users.