Service demand dynamic prediction method and system based on user classification and deep learning

Abstract

The invention provides a service demand dynamic prediction method and system based on user classification and deep learning, and the method comprises the steps: firstly carrying out the classificationof large-scale users based on a k-means algorithm, carrying out the service demand based on the service use data of each type of users, and solving problems of data sparseness and cold start; secondly, constructing a deep interactive neural network model with an enhanced attention mechanism, and carrying out the service demand prediction based on an AMEDIN model, wherein the AMEDIN model capturesinteractive relationships between different scenes and service demands in a self-adaptive mode through an interactive unit, and therefore the influences of the different scenes on the service demandsare modeled in an explicit mode; then, combining the scene features, the interaction relationship and the service demand features, and obtaining influence weights of different scenes on the service demand based on an attention mechanism; and finally, training an AMEDIN model based on the user features, the weighted scene features and the service demand features, and realizing context-aware service demand dynamic prediction based on the trained AMEDIN model.

Description

technical field [0001] The present disclosure relates to the field of computer application technology, in particular to a method and system for dynamic forecasting of service demand based on user classification and deep learning. Background technique [0002] In recent years, with the rapid development and popularization of service computing, Internet of Things, smart terminals, and 5G networks, more and more users can access feature-rich services from different fields anytime, anywhere to complete work and daily affairs. With the rapid increase in the number of available services on the network, it is difficult for users to quickly and timely find services that meet their needs, which seriously affects user satisfaction and reduces the utilization of service resources. Active service recommendation has gradually become a key technology for realizing intelligent services, and service demand dynamic prediction is the basis for realizing active service recommendation. How to ...

AI Technical Summary

Problems solved by technology

The inventor found that, on the one hand, in practical applications, it is difficult to predict service demand based on the data of a single user due to the lack of records of a single user’s use of services; most of the existing research work is based on collaborative filtering, support vector machines, matrix decomposition and machine learning methods to realize the prediction of user service needs; although the above research work has achieved good results, the accuracy of the existing methods is heavily dependent on the amount of data in the training set, and there are often data sparsity and cold start problems, resulting in a single On the other hand, the existing research work usually regards the impact of different scenarios on user service demand as equally important, which makes the model unable to fully learn the impact of different scenarios on service demand, thus The accuracy of service demand prediction is reduced; at the same time, the existing research work does not fully consider the impact of the user's scenario on its service demand, resulting in low service demand prediction accuracy

Images