Microblog Popularity Prediction Method Based on Active Learning

Abstract

The present invention provides a microblog popularity prediction method based on active learning, comprising the steps of: utilizing Sina microblog API to crawl related microblog data sets; Feature extraction of the data to obtain feature vectors; according to the extracted feature vectors, an improved model based on active learning of support vector machines is trained, and according to the trained multi-classification model, an unlabeled sample set with both uncertainty and diversity is selected. Representative samples; the selected samples are called information vectors, and handed over to experts for labeling; the labeled training data is added to the initial training set L, and this process is repeated until the performance of the model reaches a stable state to obtain Weibo popularity prediction Model. The invention reduces the problem of redundancy and the problem of outliers, reduces the number of marks of training samples, and at the same time obtains a good prediction effect in the case of less training data.

Description

technical field [0001] The invention relates to a prediction method, in particular to a microblog popularity prediction method based on active learning, which belongs to the field of machine learning. Background technique [0002] Microblog is a typical representative of social network, and it is a way for people to obtain, share and exchange information. The emergence of microblog is quietly changing the lives of modern people. Weibo attracts the attention and use of a large number of users and mass media platforms. Weibo users share information with their fans by forwarding other people's Weibo, and the fans of this user can continue to share information by continuing to forward Weibo. It also enables messages to be disseminated in large quantities and quickly through the Weibo platform to achieve information sharing. Through the Weibo platform, people can socialize with people who are far away from their own circle of life and are interested in them, and can express thei...

AI Technical Summary

Problems solved by technology

However, traditional machine learning methods also have great limitations. It requires a large amount of manually labeled data sets for model training, which requires a lot of cost, time and human resources to obtain labeled data sets.

Images