Data Prediction System Using Triple Moving Average and Rounding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing data prediction technologies fail to accurately and reliably predict future events, especially when recent trends and low-frequency events are involved, leading to inefficient resource allocation and waste in proactive response systems.
Innovation Solution
A data prediction and proactive request system using a triple moving average method that combines location-specific and item-type components, along with an improved rounding process to transform non-integer prediction values into interpretable integers, enhancing prediction accuracy and reliability.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If previous data prediction technology uses events from the same time period the previous year to predict future events, then the prediction process is simple, but the prediction accuracy deteriorates because recent trends and changes are not captured
Solution Approach 1:
The patent segments the prediction process into multiple independent components: location-specific moving averages, item-type moving averages, and their interactions. This allows each component to process specific aspects of the data independently, improving accuracy without requiring complete redesign of the entire prediction system.
Solution Approach 2:
The patent introduces additional dimensions to the prediction model by incorporating both location-specific patterns and item-type patterns simultaneously. Instead of using only temporal patterns from the same time period, the system adds spatial (location) and categorical (item-type) dimensions to capture more comprehensive trends.
2Ease of manufacture
If previous data prediction technology relies on traditional prediction methods, then the system is easy to implement, but it cannot reliably predict low-frequency or intermittent events
Solution Approach 1:
The patent changes the parameters used in prediction by switching from simple count-based predictions to probability-based predictions using moving averages. This parameter transformation allows the system to handle low-frequency events more reliably by focusing on the proportion and pattern of occurrences rather than absolute counts.
Solution Approach 2:
The system incorporates feedback mechanisms through continuous updating of moving averages based on actual observed events. This feedback loop allows the prediction model to adapt to changing patterns in low-frequency events, improving reliability over time while maintaining implementation simplicity.
3Device complexity
If inaccurate predictions are used for proactive response, then resource allocation is simplified, but resource waste increases due to incorrect item requests and transportation
Solution Approach 1:
The patent enables preliminary action by providing accurate predictions in advance, allowing the system to prepare resource allocation plans before events occur. This advance preparation ensures that resources are allocated optimally when needed, reducing waste from reactive adjustments and multiple transportation cycles.
Solution Approach 2:
By changing from integer-based count predictions to probability-based predictions, the system can allocate resources more precisely. This parameter change allows for optimized resource distribution that matches actual demand patterns, reducing both over-allocation and under-allocation of transportation and communication resources.
Data Source
AI summary
A data prediction subsystem receives event data indicating an amount of an item removed from each of a plurality of locations over a previous period of time. For each location, prediction data is determined using the event data. The prediction data includes, for each day over a future period of time, a non-integer value indicating an anticipated amount of the item that will be removed from the location. An improved rounding process is used to round the prediction value for each day. The resulting prediction data is used to proactively request items with improved communication and computational efficiency.