Production Line Control Using Decomposed Demand Forecasting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing demand forecasting methods, including time-series analysis and deep learning models, struggle to handle the complexity and volatility of modern market data, necessitating advanced hybrid models that integrate multiple models to enhance forecasting accuracy.
Innovation Solution
A method and apparatus using a hybrid machine learning model that decomposes time series data into eIMF and residue through EEMD, EMD, or CEEMDAN, extracts key variables with LASSO, and performs demand forecasting with LSTM, to control production lines in real-time based on forecasted demand.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If traditional time-series analysis and statistical methods are used for demand forecasting, then the forecasting process is simple and easy to implement, but the forecast accuracy deteriorates due to inability to handle irregular demand data and market volatility
Solution Approach 1:
The patent segments the demand forecasting process into multiple specialized models: EEMD/EMD/CEEMDAN for decomposition, LASSO/Elastic-net/Ridge for variable selection, and multiple ML models (XGBoost, LightGBM, CatBoost, RandomForest, SVM, Neural Networks) for prediction. Each model handles specific aspects of the forecasting problem, improving overall accuracy while maintaining manageable complexity through modular architecture
Solution Approach 2:
The patent creates a composite forecasting system by integrating multiple decomposition techniques (EEMD, EMD, CEEMDAN), multiple regularization methods (LASSO, Elastic-net, Ridge), and multiple machine learning models. This hybrid composite approach combines the strengths of different methodologies to achieve superior forecast accuracy compared to any single model
2Reliability
If deep learning models are applied to forecast product sales volume, then the model can handle complex market environments, but the model still cannot sufficiently handle the complexity and volatility of the market
Solution Approach 1:
The patent segments the complex forecasting problem into manageable components by first decomposing time series data using EEMD/EMD/CEEMDAN into intrinsic mode functions, then applying LASSO/Elastic-net/Ridge for feature selection, and finally using multiple ML models for prediction. This segmentation allows each component to specialize in handling specific aspects of market complexity and volatility
Solution Approach 2:
The patent merges multiple decomposition methods (EEMD, EMD, CEEMDAN), multiple regularization techniques (LASSO, Elastic-net, Ridge), and multiple machine learning models into a unified hybrid forecasting system. This combination creates a robust system that can sufficiently handle market complexity and volatility by leveraging the complementary strengths of each component
3Measurement precision
If existing hybrid models are used that integrate multiple models, then superior forecasting effects are achieved, but advanced hybrid models based on data science have not been proposed in various ways
Solution Approach 1:
The patent creates a universal hybrid forecasting framework that can accommodate multiple decomposition techniques (EEMD, EMD, CEEMDAN), multiple regularization methods (LASSO, Elastic-net, Ridge), and multiple machine learning models (XGBoost, LightGBM, CatBoost, RandomForest, SVM, Neural Networks). This multi-functional system provides superior forecast accuracy while offering various advanced data science-based hybrid model configurations for different forecasting scenarios
Data Source
AI summary
A method for controlling production lines based on product demand forecasting. The method includes constructing a first model that decomposes time series data into eIMF and a residual based on an EEMD algorithm, constructing a second model that extracts a key variable from the eIMF based on a LASSO algorithm, constructing a third model that performs demand forecasting by inputting the key variable into a machine learning model, inputting original data into the first model to decompose the original data into at least one eIMF and a residual, inputting the eIMF decomposed in the first model into the second model to extract the key variable, inputting the key variable extracted from the second model into the third model to forecast demand, and transmitting a control signal to the production lines, such that a production volume of at least one product is controlled based on the forecasted demand.