Dynamic Event Clustering With Next-Event Time Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional techniques fail to effectively track the evolution of user behavior in discrete time event sequences, as they either lack information on time components or provide static clustering results that do not consider time gaps, limiting their ability to predict next events and analyze user behavior dynamics.
Innovation Solution
A unified machine learning model, specifically a neural network, is trained to perform both supervised event predictions and unsupervised time-varying clustering, enabling the prediction of next event types and times of occurrence while dynamically clustering sequences, using techniques like LSTM networks to handle time information effectively.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional supervised learning models are used to predict next user actions, then prediction capability is provided, but information to track evolution of user behavior over time is not provided
Solution Approach 1:
The patent merges supervised learning for event prediction with unsupervised learning for dynamic clustering into a single unified neural network model. This combination allows the system to simultaneously predict next events with high accuracy while also tracking user behavior evolution through time-varying clusters, resolving the contradiction between prediction capability and behavior evolution tracking.
Solution Approach 2:
The unified neural network model performs multiple functions: it predicts next event types and times while simultaneously generating dynamic clustering results that capture user behavior evolution. This multi-functional approach eliminates the need for separate models and enables both prediction accuracy and behavior tracking to coexist.
2Adaptability or versatility
If unsupervised clustering techniques are applied to analyze event sequences, then clustering capability is provided, but time component and evolution of clusters are not considered
Solution Approach 1:
The patent implements dynamic time-varying clustering where cluster assignments evolve over time based on the sequence of events. Unlike static clustering methods, this approach adapts cluster structures dynamically, considering the temporal order and time gaps between events, thereby preserving time component information while providing versatile clustering capability.
Solution Approach 2:
The unified neural network adds a temporal dimension to traditional clustering by incorporating time information into the clustering process. The model processes event sequences in temporal order and generates cluster assignments that reflect evolution over time, effectively moving from static 2D clustering to dynamic temporal clustering.
3Reliability
If separate models are used for event prediction and clustering, then specialized functionality is achieved, but system complexity and resource consumption increase
Solution Approach 1:
The patent combines event prediction and clustering functionalities into a single unified neural network model with shared layers. This merging reduces system complexity by eliminating the need for multiple separate models while maintaining specialized functionality through dedicated output heads for prediction and clustering tasks.
Solution Approach 2:
The unified model achieves multi-functionality by processing event sequences through shared neural network layers that extract temporal features, then branching into separate prediction and clustering outputs. This architecture maintains specialized capabilities for both tasks while reducing overall system complexity and resource consumption.
4Productivity
If static clustering methods are employed, then computational simplicity is maintained, but ability to capture user behavior dynamics is limited
Solution Approach 1:
The patent implements dynamic clustering that adapts to user behavior changes over time. The neural network processes events in temporal sequence and generates time-varying cluster assignments that capture behavioral dynamics, maintaining computational efficiency through a unified model architecture rather than requiring multiple complex separate models.
Data Source
AI summary
A single unified machine learning model (e.g., a neural network) is trained to perform both supervised event predictions and unsupervised time-varying clustering for a sequence of events (e.g., a sequence representing a user behavior) using sequences of events for multiple users using a combined loss function. The unified model can then be used for, given a sequence of events as input, predict a next event to occur after the last event in the sequence and generate a clustering result by performing a clustering operation on the sequence of events. As part of predicting the next event, the unified model is trained to predict an event type for the next event and a time of occurrence for the next event. In certain embodiments, the unified model is a neural network comprising a recurrent neural network (RNN) such as an Long Short Term Memory (LSTM) network.