User network behavior feature extraction method and device and storage medium

Abstract

The invention provides a user network behavior feature extraction method. The method comprises the following steps: acquiring a network click sequence of each user; For each user, carrying out an iterative calculation process until an iterative termination condition is met: calculating the probability that the user is in each state at each moment and the probability that the user jumps from a first state to a second state at each moment in the current iterative process according to the network click sequence of the user; Calculating the probability that the initial state of the user is each state in the current iteration process and the statistical probability that the user jumps from the first state to the second state, and calculating the statistical probability that each user shows eachbehavior in each state in the current iteration process; And for each user, taking the statistical probability that the user jumps from the first state to the second state at the end of iteration asthe network behavior characteristic of the user. According to the invention, user-level network behavior feature extraction is realized.

Description

technical field [0001] The invention relates to the technical field of the Internet, in particular to a user network behavior feature extraction method, device, storage medium and electronic equipment. Background technique [0002] With the rapid development of Internet computing, not only more and more data are generated in people's daily life, but also the types of data are becoming more and more abundant. Among them, the data form represented by time series data has the most extensive coverage, such as the click sequence data of users browsing APP (application) on mobile phones, etc. In a wide range of artificial intelligence and machine learning application scenarios, how to effectively combine time series data to improve the model effect has great application requirements and research significance. Different from traditional independent and identically distributed data, time series data characterizes the characteristics of users' behavior habits over time. In order to ...

AI Technical Summary

Problems solved by technology

Due to the effect of long-tail data, the amount of data with a large length is very small, which means that the model cannot learn too long sequence information well; another method will use truncation to truncate long sequences into equal-length sequences (For example, divide a sequence with a length of 1000 into 100 sequences with a length of 10, and complete short sequences with a length of less than 10), and then use sequences of equal length for model training, but the overall information of the sequence data has been damaged

However, this type of method cannot describe the characteristics of each sequence data, and its usage scenarios also have relatively large limitations.

[0009] 3. Finally, the above two categories of methods focus on the overall data, modeling all user data as a statistical whole, without describing and extracting user-level specific features, and ignoring the characteristics of individual data; on the other hand , the existing methods are difficult to describe the global sequence information, often limited to local structural features, or local statistical features, and cannot reflect the information of the entire sequence of data

Therefore, it is difficult for existing methods to model rich time-series behaviors of users, thus limiting the final effect of time-series data models

Images