Person Flow Prediction Using Graph Time-Series Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing person flow prediction systems, such as those described in PTL 1, are inadequate as they fail to accurately predict changes in person flow when routes between points change due to variations in products or services over time, leading to potential inaccuracies in predictions.
Innovation Solution
A person flow prediction system that includes an acquisition unit for gathering attribute data and visitor numbers for exhibition articles, and a prediction unit that uses a prediction model generated from previous period data to forecast future person flow by incorporating graph time-series data on movement patterns and visitor numbers, allowing for dynamic updates and improved accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If prediction is based on flow of person and purchase amount between two points, then prediction can be performed using simple data, but prediction accuracy deteriorates when routes between points change over time
Solution Approach 1:
The patent applies dynamics by transitioning from a static two-point prediction model to a dynamic multi-point model that adapts to changing routes. The system dynamically adjusts prediction parameters based on actual visitor movement patterns observed over time, allowing the model to remain accurate even when exhibition layouts or visitor behaviors change. This is achieved by continuously updating the prediction model with new data and recalculating flow patterns across multiple points rather than fixed two-point paths.
Solution Approach 2:
The patent adds temporal and spatial dimensions to the prediction model. Instead of considering only two fixed points, the system introduces multiple observation points across the exhibition space and adds a time dimension by analyzing flow patterns over different periods. This multi-dimensional approach captures complex visitor behaviors and route changes that cannot be represented in a simple two-point model, thereby improving prediction accuracy without excessive complexity.
2Measurement precision
If prediction model uses historical data from previous periods, then prediction accuracy improves, but system cannot adapt to sudden changes in visitor behavior or exhibition content
Solution Approach 1:
The patent implements feedback mechanisms where actual visitor flow data from each period is fed back into the prediction model to continuously refine future predictions. The system compares predicted flow patterns with actual observed patterns, identifies deviations, and adjusts the model parameters accordingly. This feedback loop enables the system to maintain high accuracy while adapting to changing visitor behaviors or exhibition content, as the model learns from real-world outcomes and incorporates these lessons into subsequent predictions.
Solution Approach 2:
The patent applies preliminary action by pre-processing and storing historical flow data in structured formats before they are needed for prediction. The system pre-calculates various flow patterns, visitor behaviors, and statistical parameters from historical data, organizing them in ways that facilitate quick adaptation when changes occur. This preliminary preparation allows the model to rapidly respond to new conditions by combining pre-processed historical insights with current observations, rather than starting from scratch when changes are detected.
Data Source
AI summary
A prediction device f includes an acquisition unit and a prediction unit. The acquisition unit is configured to acquire attribute data pertaining to the plurality of exhibits and the number of visitors to each of the plurality of exhibits during a first period in the display area in which the plurality of exhibition articles are exhibited. The prediction unit is configured to predict a future flow of persons to the plurality of exhibits by a prediction model. The prediction model is generated using attribute data for a second period that is a period previous to the first period, graph time-series data pertaining to a change over time in a movement pattern for each of the plurality of visitors to the plurality of exhibits during the second period, the number of visitors to the exhibits, and the number of visitors to each of the plurality of exhibits.