Time-Series Forecasting With Downtime Pattern Segregation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Time-series forecasting models are adversely affected by intermittent downtime, leading to reduced accuracy and bias due to periods of zero or missing data, particularly when the downtime is recurring or seasonal, which existing methods like Croston's method fail to address effectively.
Innovation Solution
A novel process identifies recurring downtime patterns during model training, segregating downtime data into separate rules that are encoded within the model or a software program, allowing the model to be trained on non-downtime data and deployed with instructions to filter out downtime in the live environment, ensuring accurate predictions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If downtime data is included in training data, then the model trains on complete data sets, but the model exhibits bias and reduced accuracy
Solution Approach 1:
The patent extracts and removes downtime data points from the training data set before training the time-series forecasting model. This is achieved by detecting zero values in the time-series data that represent downtime periods and excluding them from the training process, thereby preventing bias while maintaining model reliability
Solution Approach 2:
The patent segments the time-series data into active periods and downtime periods by identifying zero values. This segmentation allows the model to process only relevant data during training while preserving the ability to handle complete data sets when needed
2Reliability
If downtime data is removed from training, then model bias is reduced, but the data set becomes incomplete
Solution Approach 1:
The patent performs preliminary detection and removal of downtime data before the training process begins. By identifying and excluding zero values in advance, the model is trained only on relevant data, preventing bias without requiring post-processing or data imputation
3Adaptability or versatility
If existing methods like Croston's method are used, then intermittent demand is addressed, but recurring seasonal downtime is not effectively handled
Solution Approach 1:
The patent changes the approach from statistical forecasting methods to a direct data removal strategy. By detecting zero values and removing them before training, the system adapts to both intermittent and recurring seasonal downtime patterns, improving accuracy for seasonal patterns that Croston's method fails to address
Data Source
AI summary
Provided is a system and method which generates a composite machine learning model that can filter downtime data from a time-series data signal and perform a prediction on remaining time-series data. In one example, the method may include detecting a pattern of downtime data within a time-series data signal, removing a subset of data from the time-series data based on the detected pattern of downtime and building a machine learning model to make predictions based on remaining data in the time-series data, generating segregation instructions configured to remove downtime data from a time-series data signal of a same type and to predict zero on future dates matching the downtime segregation codes, and building a composite machine learning model that includes the trained machine learning model and the segregation instructions for filtering data that is input to the trained machine learning models.