Point Process Learning for Prediction Accuracy with Limited Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing point process models face challenges in predicting future events when there is a scarcity of past event data and prior knowledge, particularly in new phenomena or scenarios where event tendencies differ from historical data.
Innovation Solution
A point process learning method that involves inputting a learning dataset, dividing the data by a prediction time observation area, and learning a model parameter for the intensity function of a point process model using a divided dataset, enabling accurate prediction of future events even with limited data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If many pieces of event data and prior knowledge are required for accurate prediction, then prediction accuracy is improved, but data availability deteriorates
Solution Approach 1:
The patent applies preliminary action by pre-dividing the learning dataset into multiple datasets with different time spans before training begins. This preprocessing step creates multiple training scenarios that teach the model to handle varying amounts of historical data, enabling accurate predictions even when actual available data is limited. The division is performed based on prediction time observation areas, creating datasets that simulate different data availability conditions.
Solution Approach 2:
The patent changes the parameter of time span in the learning dataset by creating multiple datasets with different temporal coverage (e.g., short-term, medium-term, long-term event histories). This parameter variation allows the model to learn how to adapt its predictions based on the amount of available historical data, improving prediction accuracy across different data availability scenarios without requiring actual increases in data quantity.
2Measurement precision
If prior knowledge is incorporated into the model, then prediction accuracy is improved, but model complexity increases
Solution Approach 1:
The patent segments the learning process by dividing the learning dataset into multiple distinct datasets based on time span characteristics. Each segmented dataset is used to train or fine-tune the model for specific prediction scenarios. This segmentation allows the model to specialize in different temporal patterns without requiring a single complex model structure that tries to handle all scenarios simultaneously, thereby reducing overall model complexity while maintaining or improving prediction accuracy.
3Adaptability or versatility
If the model is trained on diverse event data, then adaptability to new scenarios is improved, but training data requirements increase
Solution Approach 1:
The patent performs preliminary division of the learning dataset into multiple datasets with different time spans before training begins. This preprocessing creates a structured collection of training scenarios that cover diverse temporal patterns. When the model is trained on this pre-divided dataset collection, it learns to adapt to different time span characteristics and new scenarios without requiring separate training data for each scenario, thereby improving adaptability while controlling training data requirements through efficient data reuse across different time span divisions.
Data Source
AI summary
According to an embodiment, a point process learning method executed by a computer includes: an input procedure of inputting a learning data set including at least first event data representing a series of occurrences of first events; a division procedure of dividing the first event data included in the learning data set by using a prediction time observation area including at least a time series when predicting future event occurrence; and a learning procedure of learning a model parameter including a parameter of an intensity function of a predetermined point process model by using a divided learning data set divided in the division procedure.