ML Resource Forecasting for Dynamic Enterprise Workloads
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional resource forecasting techniques often produce inaccurate and erroneous results due to reliance on individual signals and static rules, failing to account for the influence of dynamic changes and externalities.
Innovation Solution
Implementing machine learning techniques to correlate resource-related data with target variables, using methods such as random forests and linear regression models, to generate forecasts and automate actions based on these forecasts.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If conventional resource forecasting techniques use individual signals and static rules, then the forecasting process is simple and easy to implement, but the forecasting accuracy deteriorates and produces inaccurate results
Solution Approach 1:
The patent applies dynamics by transitioning from static forecasting rules to dynamic machine learning models that continuously adapt to changing resource consumption patterns. The system uses historical data to train models that can dynamically adjust predictions based on new inputs, capturing temporal variations and complex relationships that static rules cannot handle.
Solution Approach 2:
The patent changes parameters by moving from fixed threshold-based forecasting to probabilistic predictions generated by machine learning models. The system outputs forecasted values with associated confidence intervals, allowing for more nuanced decision-making. Multiple machine learning algorithms (e.g., random forests, gradient boosting) with different parameters are employed to optimize forecasting accuracy for specific resource types.
2Device complexity
If conventional techniques rely on static rules, then the system complexity is low, but the ability to account for dynamic changes and externalities deteriorates
Solution Approach 1:
The patent applies preliminary action by training machine learning models on historical resource consumption data before actual forecasting is needed. This pre-training phase allows the system to learn patterns and relationships in advance, so that when actual forecasting is required, the models can quickly adapt to new conditions without requiring complex real-time adjustments.
Solution Approach 2:
The patent implements feedback mechanisms where forecasted values are compared against actual resource consumption, and this information is used to retrain and refine the machine learning models. This continuous feedback loop enables the system to adapt to changing conditions and improve accuracy over time, capturing dynamic changes that static rules cannot accommodate.
3Measurement precision
If machine learning techniques are used to correlate resource data, then forecasting accuracy is improved, but the computational complexity and data processing requirements increase
Solution Approach 1:
The patent applies segmentation by dividing the forecasting problem into separate machine learning models for different resource types (CPU, memory, storage, network). Each model is trained on specific resource data and can be independently optimized. This segmentation reduces the overall computational complexity compared to a single monolithic model while maintaining high accuracy for each specific resource type.
Data Source
AI summary
Methods, apparatus, and processor-readable storage media for resource-related forecasting using machine learning techniques are provided herein. An example computer-implemented method includes obtaining multiple items of data related to one or more resources associated with an enterprise; correlating at least a portion of the multiple items of data with at least one target variable using one or more correlation techniques; generating one or more forecasts pertaining to the at least one target variable and at least a portion of the one or more resources by processing at least a portion of the correlated data using one or more machine learning techniques; and performing one or more automated actions based at least in part on the one or more forecasts.