4 results about "Internet users" patented technology

This invention belongs to the field of internet user preference prediction technology, specifically relating to a user preference prediction method based on data debiasing. The method includes: acquiring video data and user data; processing the video data and user data using an embedding layer to obtain user embeddings and video embeddings; debiasing the user embeddings and video embeddings to obtain debiased user embeddings and debiased video embeddings; processing the user embeddings, video embeddings, debiased user embeddings, and debiased video embeddings using a feature filtering layer to obtain user-specific feature importance weights and video-specific feature importance weights; processing the user-specific feature importance weights and video-specific feature importance weights using a factorization machine to obtain a user interest representation vector; and inputting the user interest representation vector into an output layer for processing to obtain the user preference prediction result. This invention can accurately and effectively predict user preferences, improving the accuracy of user preference prediction.

This invention discloses an intranet traversal system, comprising a client and a server. The client runs on a computer within a local area network (LAN), and the server runs on a cloud host. Each client connects to an internal service within the LAN. When the corresponding internal service requires traversal, the client initiates a connection request to the server. The server receives the connection request from the client, establishes a tunnel with the client, and starts a proxy service for the internal service corresponding to that client. The proxy service receives access requests from internet users and forwards them to the internal service within the LAN via the server and client. This invention provides strong computing and storage services while controlling costs and simplifying intranet traversal operations.

This invention belongs to the field of internet user access experience technology, specifically relating to an internet access quality monitoring system and method for real-time perception of user experience. This invention constructs a diffusion intensity index by using the adjacency matrix change rate and the experience quality index change rate. This index can identify whether network quality anomalies are propagating in the spatial structure, achieving an improvement from single-point anomaly monitoring to group trend monitoring. The network performance sampling frequency is dynamically adjusted according to the diffusion intensity, improving sampling accuracy during anomaly propagation and reducing sampling load in stable conditions. This reduces system resource consumption while ensuring monitoring accuracy. By integrating individual network performance, spatial adjacency relationships, time window fluctuation statistics, and diffusion trend judgment, real-time dynamic perception and adaptive monitoring and control of user experience quality are achieved, resulting in significant improvements in anomaly identification accuracy, group diffusion perception capability, and system resource utilization efficiency.

This invention relates to a method for predicting internet user ad clicks based on big data, belonging to the field of internet advertising technology. It aims to solve the problems of difficulty in modeling high-dimensional sparse features, insufficient capture of feature interactions, and weak prediction generalization ability in existing technologies. The method first acquires user features, ad features, and historical behavior sequences. User and ad features are divided into coarse-grained and fine-grained categories and encoded. Low-dimensional dense features are extracted through feature processing. Next, the user's historical behavior sequences are embedded and sequence modeled to obtain dynamic interest vectors. Subsequently, the dynamic interest vectors are concatenated with the coarse-grained low-dimensional features, and dynamic sample clusters are obtained through clustering. A dedicated model is built for each cluster, fusing fine-grained features. Finally, the fused features are input into a deep neural network containing both click and deep interaction tasks, outputting the prediction results. This invention, through layered processing of coarse and fine-grained features, clustering and grouping modeling, and dual-task collaborative optimization, effectively mines deep feature correlations, accurately captures user dynamic interests, and significantly improves the accuracy and generalization ability of click prediction, providing reliable support for precise ad targeting.