Hybrid Correction for Time Series Forecasting with Noisy Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for estimating impression counts for advertisements, such as those on billboards, face inaccuracies due to inconsistent smartphone location data, leading to the need for correction schemes that can adapt to different data characteristics and patterns.
Innovation Solution
A hybrid approach is implemented, using a forecasting time series model to estimate impression counts and applying dynamic correction based on profile metrics, which detects specific data characteristics and applies appropriate correction schemes to improve the accuracy of impression counts.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If smartphone location data is used to estimate impression counts, then the estimation can be performed, but the data is inconsistent and inaccurate
Solution Approach 1:
The system uses profile metrics to detect data characteristics and applies feedback loops to adjust correction schemes dynamically. The correction scheme continuously monitors the relationship between measured and estimated impression counts, adapting to changing data patterns and improving accuracy over time through iterative refinement.
Solution Approach 2:
The system changes parameters by switching between different correction schemes based on detected data characteristics. When specific patterns are identified through profile metrics, the system adjusts correction parameters and selects appropriate correction strategies, transforming the rigid estimation process into a flexible, adaptive system.
2Measurement precision
If correction schemes are applied to improve accuracy, then impression count estimation improves, but the system complexity increases
Solution Approach 1:
The correction system is segmented into distinct modules: profile metrics computation, data characteristic detection, and correction scheme application. Each module handles a specific aspect of the correction process independently, making the overall complex system more manageable and maintainable through functional decomposition.
Solution Approach 2:
Profile metrics serve as an intermediary layer between the raw measured impression counts and the correction schemes. Rather than directly applying complex corrections to raw data, the system first computes profile metrics that characterize data patterns, then uses these metrics to select and apply appropriate correction schemes, simplifying the correction process.
3Adaptability or versatility
If dynamic correction based on profile metrics is applied, then adaptability to data characteristics improves, but processing time increases
Solution Approach 1:
The system performs preliminary computation of profile metrics and detection of data characteristics before applying correction schemes. By preparing the correction strategy in advance based on detected patterns, the actual correction process becomes more efficient and less time-consuming, as the correction logic is already determined.
Solution Approach 2:
The system dynamically adjusts the correction process based on real-time detection of data characteristics through profile metrics. Rather than using a fixed, time-consuming correction process, the system adapts its correction strategy based on the actual data patterns detected, optimizing processing time for each specific data scenario.
Data Source
AI summary
The present teaching relates to impression count determination. A forecasting time series (TS) model is established based on measured impression counts (MI-counts). Metrics are calculated from MI-counts from a sub-range of a profile. Different types of data characteristics are detected based on the metrics. Hybrid correction applied to the profile is determined based on the detected data characteristics. Corrected impression counts (I-counts) for the profile are generated via the hybrid correction operation based on the MI-counts and I-counts estimated from the forecasting TS models and provided for determining a level of viewership of the content at the site.