Ensemble Learning for Multi-Variable Time-Series Forecasting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current machine learning techniques are limited in predicting future time-series data, as they often focus on single feature predictions for a single time period, failing to account for multiple variables and multiple time periods, which restricts their application in downstream applications.
Innovation Solution
An ensemble model is developed that includes a feature engineering pipeline to transform raw data into feature vectors, using a combination of intrinsic and external data augmentation, and employs a predictive model (such as LightGBM or LSTM) along with a meta-model for generating predictions across multiple time periods, integrating predictions from multiple models for improved accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If most techniques predict only a single feature for a single future time period, then the prediction model remains simple, but the application scope is limited and cannot support downstream applications requiring multiple variables and time periods
Solution Approach 1:
The patent segments the prediction task into multiple independent prediction heads, where each head is responsible for predicting a specific target variable. This allows the model to handle multiple variables simultaneously while maintaining modular architecture, thus improving adaptability without proportionally increasing overall complexity.
Solution Approach 2:
The patent extends the prediction capability from a single time point to multiple future time periods by adding a time dimension. The model predicts sequences of future values rather than single points, enabling applications that require temporal forecasting across multiple periods while using structured approaches to manage the increased complexity.
2Reliability
If an ensemble model with multiple base models and meta-model is used to predict across multiple variables and time periods, then prediction accuracy and reliability are improved, but the model complexity increases
Solution Approach 1:
The patent combines multiple base models (including traditional machine learning models and deep learning models) into an ensemble system where their predictions are aggregated. This merging of diverse modeling approaches improves prediction reliability by leveraging the strengths of different algorithms while managing complexity through systematic integration.
Solution Approach 2:
The patent introduces a meta-model as an intermediary layer that processes and integrates the predictions from multiple base models. This meta-model serves as a mediator that synthesizes outputs from different sources, improving overall prediction reliability while providing a structured approach to managing the complexity of the ensemble system.
Data Source
AI summary
The disclosure relates to predicting the future value of a time series. In an embodiment, a method is disclosed which includes generating a feature vector, the feature vector comprising a set of raw features and a plurality of lag features, the plurality of lag features including a current value of a selected feature in the set of raw features and one or more historical values of the selected feature; inputting the feature vector into a plurality of base models, the plurality of base models outputting a plurality of predictions, each prediction in the plurality of predictions representing future values of the selected feature; and predicting a future value of the selected feature by inputting the plurality of predictions into a meta-model.