Event prediction in social network through Twitter messages analysis

Area of Science:

• Computational Social Science
• Data Mining
• Natural Language Processing

Background:

• Social media analysis is increasingly important for event detection.
• Predicting events, particularly during social crises, holds significant value.
• Existing methods require refinement for accurate and timely event prediction.

Purpose of the Study:

• To develop a novel method for predicting future events using Twitter data.
• To analyze changes in tweet rates to identify precursors to events.
• To enhance the precision of event prediction through social media analysis.

Main Methods:

• Tweets were preprocessed within fixed-length time windows.
• Non-negative matrix factorization and distance-dependent Chinese restaurant process incremental clustering were used for tweet categorization.
• Event prediction was based on identifying significant increases in tweet volume within clusters.

Main Results:

• A high rate of tweets entering a cluster indicates a potential future event.
• Investigations over a 6-month period showed considerable changes in tweet rates preceding predictable events.
• The proposed method demonstrated the capability for high-precision event prediction.

Conclusions:

• Analyzing tweet rate fluctuations is a viable strategy for event prediction.
• The developed method offers a promising approach for early event detection, especially for social crises.
• This technique can significantly improve the accuracy and timeliness of event forecasting.