The application discloses a kind of virtual avatar construction and memory iteration method based on multi-dimension user characteristics, it is related to virtual avatar technical field, including step one: real-time collection text data, and pre-
processing;Step two: the module of distinguishing vector is calculated to the text data after pre-
processing, and user portrait is constructed according to the module of distinguishing vector;Step three: according to the running law of user portrait, the running weight is calculated, and memory iteration is carried out according to the running weight.The application first quantifies
semantics accurately by calculating the module of text distinguishing vector, and realizes the
dynamic clustering and updating of user portrait according to the module value matching;Further, the running weight of each portrait is calculated using the
Euclidean distance formula by combining the call frequency and duration;Finally, based on
bubble sort and median method, high-weight memory is retained and iterated, so as to realize the continuous evolution and long-term personalized accompaniment of virtual avatar in
limited resources.