Adaptive Time Series Prediction Model Generation via Stratified Sampling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional predictive data models are not adaptive to specific data sets, leading to inaccurate predictions due to static generation and assumptions that do not align with actual data patterns, resulting in less effective forecasting.
Innovation Solution
Generating prediction models specific to observed time series data sets by selecting parameter values using stratified sampling and efficient computation methods, such as ARIMA models, to improve accuracy and adaptability.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional predictive data models are generated statically using general assumptions, then the model generation process is simple and fast, but the prediction accuracy deteriorates because the models do not adapt to specific data set patterns
Solution Approach 1:
The patent transforms static predictive models into dynamic adaptive models that automatically adjust their parameters based on the specific characteristics of each data set. The system dynamically selects and tunes model parameters (such as ARIMA orders) to match the temporal patterns, seasonality, and trends of the input data, enabling the model to adapt its structure rather than remaining fixed.
Solution Approach 2:
The patent systematically varies and optimizes model parameters including autoregressive order (p), moving average order (q), and differencing order (d) in ARIMA models. By changing these parameters based on data set characteristics and evaluating their impact on prediction accuracy, the system finds optimal parameter configurations that improve predictive performance for each specific data set.
2Measurement precision
If prediction models are customized for each specific data set, then prediction accuracy improves, but the computational time and resources required increase significantly
Solution Approach 1:
The patent performs preliminary analysis of data set characteristics (such as stationarity testing, seasonality detection, and trend identification) before full model training. This preliminary action includes pre-processing steps like differencing to achieve stationarity and identifying appropriate model types, which streamlines the subsequent parameter optimization process and reduces overall computation time.
Solution Approach 2:
The patent implements feedback mechanisms where model predictions are continuously evaluated against actual data, and the results are used to refine parameter selections. This feedback loop allows the system to learn from prediction errors and adjust parameters iteratively, improving accuracy while avoiding exhaustive search of all possible parameter combinations.
3Reliability
If general predictive models are used across different data sets, then the ease of operation is high, but the reliability of predictions deteriorates due to mismatch with specific data patterns
Solution Approach 1:
The patent enables the predictive model system to automatically analyze data characteristics, select appropriate model types, and optimize parameters without requiring manual intervention or expert knowledge. The system self-configures by detecting data patterns, selecting relevant features, and tuning hyperparameters automatically, making reliable customized modeling accessible without increasing operational complexity for users.
Data Source
AI summary
Embodiments of the present invention provide systems, methods, and computer storage media for facilitating generation of prediction models. In some embodiments, a predetermined number of parameter value sets is identified. Each parameter value set includes a plurality of parameter values that represent corresponding parameters within a time series model. The parameter values can be selected in accordance with stratified sampling to increase a likelihood of prediction accuracy. The parameter value sets are input into a time series model to generate a prediction value in accordance with observed time series data, and the parameter value set resulting in a least amount of prediction error can be selected and used to generate a time series prediction model (ARIMA, AR, MA, ARMA) with corresponding model parameters, such as p, q, and/or k, subsequently used to predict values.