Multi-variable time series forecasting is one of several applications of machine learning. Creating an artificial environment capable of replicating real-world behavior is useful for understanding the intrinsic relationship between variables. However, selecting a predictor that ensures good performance for variables of different natures is not always a simple process. An algorithmic approach based on metaheuristics could be a good alternative to find the best predictive model for variables. Each predictor is optimized for forecasting a particular variable in a multi-agent artificial environment to improve the overall performance. The resulting environment is compared with other solutions that use only the same type of predictor for each variable. Finally, we can assert that using a multi-agent environment can improve the performance, accuracy, and generalization of our model.
Optimizing Multi-variable Time Series Forecasting Using Metaheuristics
Zito F.;Cutello V.;Pavone M.
2023-01-01
Abstract
Multi-variable time series forecasting is one of several applications of machine learning. Creating an artificial environment capable of replicating real-world behavior is useful for understanding the intrinsic relationship between variables. However, selecting a predictor that ensures good performance for variables of different natures is not always a simple process. An algorithmic approach based on metaheuristics could be a good alternative to find the best predictive model for variables. Each predictor is optimized for forecasting a particular variable in a multi-agent artificial environment to improve the overall performance. The resulting environment is compared with other solutions that use only the same type of predictor for each variable. Finally, we can assert that using a multi-agent environment can improve the performance, accuracy, and generalization of our model.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.