E-commerce consumption path prediction and loss early warning system based on multi-modal fusion

By fusing the Transformer-XL model with multimodal data, the difficulty of identifying the state transition phase in e-commerce user behavior analysis was solved, enabling accurate prediction of user states and timely adjustment of recommendation signals, thereby improving the operational efficiency and user experience of e-commerce platforms.

CN121329488APending Publication Date: 2026-01-13SHANDONG QINGQIAO INFORMATION TECHNOLOGY CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202511501042.8
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2025-10-21
Publication Date
2026-01-13

AI Technical Summary

Technical Problem

In existing e-commerce user behavior analysis and decision path modeling, the lack of effective integration of multimodal data leads to difficulties in identifying the transition phase of user state transformation, vague or delayed user state prediction, inaccurate recommendation signal output, biased consumption path construction, and untimely churn warning.

Method used

The Transformer-XL model is used to train sub-units. The model is trained by multimodal fusion feature sequences. Combined with e-commerce log processing and segmentation length optimization sub-units, a multimodal fusion modeling and state prediction module, a path construction and recommendation signal generation module, and a state optimization and recommendation signal adjustment module are constructed to achieve effective fusion of multimodal data and state prediction.

Benefits of technology

It improved the accuracy of user status prediction, optimized the construction of consumption paths, enhanced the timeliness and accuracy of recommendation signals, reduced the risk of user churn, and improved the operational efficiency and user experience of e-commerce platforms.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN121329488A_ABST
    Figure CN121329488A_ABST
Patent Text Reader

Abstract

The invention relates to the technical field of e-commerce data processing, in particular to a multi-modal fusion-based e-commerce consumption path prediction and loss early warning system, which comprises a multi-modal fusion modeling and state prediction module, a path construction and recommendation signal generation module and a state optimization and recommendation signal adjustment module, the multi-modal fusion modeling and state prediction module is used for training a Transform-XL model according to the multi-modal fusion feature sequence; the Transform-XL model outputs the prediction probability of each state, and the state with the maximum prediction probability is defined as an output state; the path construction and recommendation signal generation module divides the length to construct a plurality of full-chain paths, and analyzes key operations in the plurality of full-chain paths to construct a consumption path; and finally generating recommendation signal output according to the consumption path. And if the change trend of the probability is predicted in each state, the state optimization and recommendation signal adjustment module fuses a plurality of consumption paths in the tendency state, inputs the fused consumption paths to the Transform-XL model, and adjusts the output state.
Need to check novelty before this filing date? Find Prior Art