SIR Model Virality Prediction for Social Media

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Social media platforms lack effective methods to predict whether a post will go viral, leading to the spread of false or misleading information that can be detrimental, as existing approaches often rely on content analysis and may result in false positives due to limited cluster analysis.

Innovation Solution

A computer-implemented method using a susceptible-infected-recovered (SIR) model as a differential equation to analyze social media items, linking them to breakout patterns based on predefined dimensions, bucketing them for similarity, and determining virality through a gamma depth walk and Trie structured datasets, identifying items with high propensity for spread.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If content analysis methods are used to predict viral spread, then the prediction process can be performed, but false positives occur due to limited cluster analysis

Engineering Contradiction:
Improveprediction accuracyVSAvoidfalse positive rate
Core Design Contradiction:
Measurement precisionVSReliability

Solution Approach 1:

The patent segments the social media item universe into distinct buckets based on breakout pattern similarity. Each bucket represents a different virality trajectory, allowing the system to analyze items within homogeneous groups rather than treating all items uniformly. This segmentation enables more accurate predictions by comparing items against relevant historical patterns specific to their bucket category.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent introduces a new dimensional approach by applying the SIR (Susceptible-Infected-Recovered) epidemiological model to social media virality prediction. This transforms the prediction problem from traditional content analysis into a dynamic systems model that tracks the spread of information through population states, adding a temporal and behavioral dimension to the analysis that captures actual user interaction patterns.

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

2Adaptability or versatility

If traditional cluster analysis is used, then items can be grouped, but the analysis is limited and cannot accurately identify breakout points

Engineering Contradiction:
Improvepattern recognition capabilityVSAvoidbreakout point identification accuracy
Core Design Contradiction:
Adaptability or versatilityVSMeasurement precision

Solution Approach 1:

The patent changes the parameters used for grouping items from traditional content-based clustering to similarity metrics based on breakout patterns. Instead of grouping by topic or content type, items are bucketed based on their similarity to historical breakout trajectories, using parameters such as engagement velocity, share patterns, and temporal dynamics. This parameter transformation enables the system to identify items likely to break out regardless of their content category.

Inventive Principle:
Principle #35Parameter changes

Solution Approach 2:

The system incorporates feedback loops where predicted viral items are monitored to validate predictions. When items are identified as likely to go viral based on SIR model predictions, the system tracks their actual performance and uses this feedback to refine the breakout pattern recognition and bucketing algorithms, continuously improving breakout point identification accuracy over time.

Inventive Principle:
Principle #23Feedback

Data Source

PatentUS11494439B2Digital modeling and prediction for spreading digital data
Publication Date: 2022.11.08 INTERNATIONAL BUSINESS MACHINE CORPORATION
  • US11494439B2 patent drawing
  • US11494439B2 patent drawing
  • US11494439B2 patent drawing

AI summary

A method, computer program product, and a system where a processor(s) obtains a universe of social media items posted to a social media platform at a given time. The processor(s) determines a likelihood of each item of the universe of social media items becoming viral by linking each item of the universe to a breakout pattern based on predefined dimensions and bucketing each item into a defined grouping by a measure of anticipated virality of the item. The processor(s) applies a susceptible-infected-recovered (SIR) model as a differential equation to the bucketed items to determine an equilibrium and one or more breakout points. The processor(s) determines if a social media item will become viral based on whether the item matches at least one or the one or more breakout points.