Traffic Prediction Grouping by Communication Quality
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing traffic prediction methods fail to accurately predict traffic based on required communication quality, as they group traffic solely by time-series fluctuations, leading to unstable fluctuations and decreased prediction accuracy.
Innovation Solution
A prediction apparatus that calculates statistics for traffic combinations of users, services, and time zones, extracts patterns, classifies users and services into groups based on similarity, and applies prediction techniques to each group to improve accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If traffic is divided in units of services and users for prediction, then traffic prediction can be performed for each service type, but each traffic volume becomes small causing unstable fluctuations and decreased prediction accuracy
Solution Approach 1:
The patent merges multiple traffic data dimensions (service type, terminal environment, user characteristics) into unified traffic groups. By combining these different attributes, the system creates sufficiently large traffic volumes for stable statistical analysis while preserving service-specific prediction capabilities through the hierarchical group structure.
2Ease of operation
If traffic is grouped solely by time-series fluctuations, then prediction ease is improved, but the grouped traffic contains services with different required communication qualities making quality-based path control impossible
Solution Approach 1:
The patent applies local quality by creating different grouping dimensions for different purposes: time-series-based grouping for prediction ease, and service/terminal environment-based grouping for communication quality control. Each grouping method serves its specific function while maintaining overall system versatility.
Solution Approach 2:
The patent adds new grouping dimensions (service type, terminal environment, user characteristics) beyond traditional time-series grouping. This multi-dimensional approach allows simultaneous achievement of prediction ease through temporal grouping and quality-based control through service/terminal grouping.
3Adaptability or versatility
If traffic volume is small for each service and user combination, then detailed service-specific analysis is possible, but traffic fluctuations become unstable making trend determination difficult
Solution Approach 1:
The patent combines traffic data from multiple services and terminal environments into unified groups to achieve stable traffic volumes. This merging maintains service-specific analysis capability through the hierarchical structure while ensuring sufficient data volume for stable statistical analysis and trend determination.
Data Source
AI summary
A prediction apparatus includes, a first calculation unit configured to calculate, for traffic in a past time period between a plurality of users and a plurality of services, statistics of the traffic in units of combinations of the users, the services, and time zones, an extraction unit configured to extract, based on the statistics, a plurality of patterns from the traffic, a classification unit configured to classify, by pattern, at least one of each user or each service into groups, a second calculation unit configured to calculate, based on the classification result from the classification unit and the statistics, per-time-zone statistics for each group, and a third calculation unit configured to apply a prediction technique to the per-time-zone statistics for each group, to calculate, for each group, a prediction value of the statistics for a time period later than the past time period, resulting improved accuracy of traffic prediction.