Looking for breakthrough ideas for innovation challenges? Try Patsnap Eureka!

Event sequence prediction method based on time sequence convolution and relation modeling

A technology of event sequences and prediction methods, applied in prediction, neural learning methods, character and pattern recognition, etc., can solve problems such as low computational efficiency, limited model performance, no event mining and analysis, etc., to achieve rich expression ability and improve prediction The effect of improving performance and computing efficiency

Pending Publication Date: 2022-07-08
NANJING UNIV
View PDF0 Cites 3 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0006] Although the model based on the cyclic neural network already has a certain prediction accuracy, due to the characteristics of the cyclic neural network's own structure, it is difficult to perform parallel computing, so there is a problem of low computational efficiency in the process of model training and reasoning.
Moreover, although the cyclic neural network can encode and express sequence data, it cannot directly describe the influence relationship between different events in the sequence. Therefore, the correlation model not only lacks interpretability, but also does not fully understand the influence relationship between events. mining and analysis, limiting model performance

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Event sequence prediction method based on time sequence convolution and relation modeling
  • Event sequence prediction method based on time sequence convolution and relation modeling
  • Event sequence prediction method based on time sequence convolution and relation modeling

Examples

Experimental program
Comparison scheme
Effect test

Embodiment

[0099] In order to verify the effectiveness of the present invention, we conduct example verification on two event sequence prediction task datasets collected in real scenarios, including IPTV dataset and MIMIC-II dataset. Among them, the IPTV dataset is the event sequence data provided by China Telecom Corporation on the behavior of users watching cable TV programs, including the viewing behavior sequences of 2,967 users; the MIMIC-II dataset is the medical diagnosis-related event sequence data of patients provided by the medical center. Visit data for 53,423 patients who visited the medical center between 2001 and 2008 were recorded. The present embodiment now takes an event sequence data for testing in the IPTV data set as an example, and predicts events according to the following steps:

[0100] 1. Perform data preprocessing on the event sequence data, including data cleaning and calculation of interval time. The data cleaning process includes the removal of invalid sampl...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

PUM

No PUM Login to View More

Abstract

The invention discloses an event sequence prediction method based on time sequence convolution and relation modeling. The event sequence prediction method comprises the following steps: step 1, extracting an event sequence training set from a database; 2, preprocessing the original data; step 3, performing feature extraction on the mark information in the historical sequence by using a mark feature encoder; 4, performing feature extraction on the time information in the historical sequence by using a time sequence feature encoder; step 5; performing feature fusion on the mark feature codes and the time feature codes of the historical events, and outputting feature representation of a single historical event; step 6, constructing a time sequence correlation graph between events on the basis of the event codes, and outputting feature codes of a historical sequence; 7, respectively calculating condition intensity for each type of event; 8, calculating a model loss function and updating parameters; step 9, judging whether the loss curve of the model is converged or not, and if not, returning to step 8; and step 10, storing the trained model and performing deployment.

Description

technical field [0001] The invention relates to an event sequence prediction method, in particular to an event sequence prediction method based on time series convolution and relational modeling. Background technique [0002] The problem of event sequence prediction is an important research direction in the field of time series analysis. As early as the 1970s, some related scholars carried out related research work in this field. In our daily life, event sequence data is ubiquitous, and technical achievements related to event sequence prediction are applied in many scientific fields, such as social sciences, medicine, geology, and physics. It is of great social value to analyze and understand data related to event sequences and to make accurate predictions of future events, so this field has received extensive attention from the academic community. [0003] Conventional event sequence data is a set of sequences arranged according to the sequence of events. Each event sample...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

Application Information

Patent Timeline
no application Login to View More
Patent Type & Authority Applications(China)
IPC IPC(8): G06N3/04G06N3/08G06K9/62G06Q10/04
CPCG06N3/049G06N3/084G06Q10/04G06N3/045G06F18/253
Inventor 申富饶王言赵健
Owner NANJING UNIV
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Patsnap Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Patsnap Eureka Blog
Learn More
PatSnap group products