Proactive Data Prediction System Using Triple Moving Average
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing data prediction technologies fail to accurately and reliably predict future events, especially when recent trends and changes are involved, and struggle with low-activity events that occur intermittently, leading to inefficient resource allocation and waste in responding to predicted needs.
Innovation Solution
A proactive request communication system that uses a triple moving average method combining location-specific and zone-specific components, along with an improved rounding process to transform non-integer prediction values into interpretable integers, allowing for more accurate and dynamic event predictions, even at outlier locations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If previous data prediction technology uses events from the same time period in the previous year for predictions, then historical data is utilized, but recent trends and changes are not captured
Solution Approach 1:
The patent segments the prediction model into multiple components: location-specific components (capturing local recent trends) and zone-aggregated components (capturing broader patterns). This segmentation allows the system to simultaneously utilize recent location-specific data while maintaining stability through aggregated zone data, resolving the contradiction between capturing recent trends and utilizing historical data.
Solution Approach 2:
The patent introduces a temporal dimension by using moving averages with different windows (e.g., 7-day, 14-day, 28-day averages). This multi-timescale approach allows the system to capture recent changes while smoothing out noise, enabling accurate predictions that reflect both recent trends and historical patterns without being limited to a single time period comparison.
2Adaptability or versatility
If previous data prediction technology predicts events with limited information, then predictions can be made for intermittent events, but prediction reliability is low
Solution Approach 1:
The patent merges location-specific moving averages with zone-aggregated moving averages to create a composite prediction model. For intermittent events at a location with limited data, the zone-aggregated component provides additional information from similar locations, improving reliability while maintaining the ability to handle intermittent event patterns. This combination allows the system to make reliable predictions even when individual location data is sparse.
3Ease of operation
If prediction values are rounded to integers, then results are interpretable, but rounding errors accumulate
Solution Approach 1:
The patent applies preliminary rounding adjustments before final integer conversion. The system first calculates precise floating-point prediction values using the moving average model, then applies a rounding algorithm that considers the cumulative effect of rounding across multiple predictions. This preliminary processing maintains precision during calculations while ensuring final interpretability, reducing cumulative rounding errors compared to naive rounding approaches.
Data Source
AI summary
A data prediction subsystem receives event data indicating amounts of items removed from locations over a previous period of time, An event probability is determined based at least in part on a number of concurrent days without detected item removal events for a first item at a first location and an anticipated item removal amount per day. After determining that the event probability is less than the threshold value, an updated status is determined for the first item at the first location. The updated status is an empty status indicating that the first item is not believed to be present at the first location. Based at least in part on the updated status for the first item at the first location, a prediction value is determined corresponding to a recommended amount of the first item to request for a future time.