Image recommendation method based on user psychological characters and progressive dissociation

The invention discloses an image recommendation method based on user psychological characters and progressive dissociation, and belongs to the field of image recommendation. According to the method, firstly, a user high-order strong and weak characteristic preference network is constructed, and deep preference representation of users is realized; then, a shopping psychology-character extraction capsule based on a sliding window network is put forward to deduce psychology and character differences of different commodities purchased by the user, and shopping behavior psychology and character features of the user are deeply represented in a multi-dimensional mode. Finally, a progressive preference dissociation method is put forward, click and purchase adding behaviors are fused, auxiliary tasks are introduced subsequently to purchase, like and use shopping psychology and characters to serve as basic tasks, auxiliary tasks are introduced subsequently to derive continuous purchase preferences, multiple preference