Time Series Anomaly Prediction Based on XGBoost: An Intelligent Early Warning System for Information Supply-Demand Imbalance

This article introduces an open-source machine learning project whose core is to use an XGBoost classifier to predict time series anomalies. By analyzing the relationship between GDELT news data (supply side) and Wikipedia pageview data (demand side), it constructs an information difference indicator and achieves early warnings two days in advance for abnormal states of "vacuum" and "surplus" in the information environment. This system aims to solve the problem of forward-looking monitoring of information ecosystem imbalance and is of great value to multiple fields such as news media and investment institutions.

Supply-demand imbalance in the information ecosystem (such as information vacuum or surplus) often indicates changes in social dynamics and affects public opinion, market decisions, etc. Traditional public opinion monitoring is mostly post-hoc analysis. The goal of this project is to achieve forward-looking prediction—issuing early warnings two days in advance to provide decision support for news media, policy makers, etc.

The project quantifies the Information Delta:

1. Data sources: GDELT global news event data for the supply side, and Wikiviews pageview data for the demand side.
2. Data standardization: Raw data is divided by the expected value to eliminate dimensional differences: $

This article introduces a machine learning project that uses an XGBoost classifier to predict time series anomalies. By analyzing the relationship between GDELT news data and Wikipedia pageview data, the system constructs an information supply-demand difference indicator and can issue early warnings two days in advance for abnormal states of "vacuum" and "surplus" in the information environment.