IntellTech presents research in the Time Series Prediction Area
Anticipating future events has been in the interest of humanity since the dawn of its existence. In the field of Geotechnical Engineering, this is no different. Anticipating unsafe conditions can allow those responsible for mitigating risks, whether performing engineering works, preventive maintenance on structures, or, in extreme cases, evacuating locations.
In this sense, many branches of science intend to build tools that allow such anticipations; one of them is the Time Series Analysis, whose objective is forecasting. Briefly, this branch of science consists of, given a sequence of observations of one or more variables taken over time, extracting relevant information from these series. In the context of forecasting, the objective is to estimate the observations yet to be taken.
IntellTech strives to make the most accurate and currently available to its partners, including that related to time series modeling. The SHMS already has tools for forecasting; however, in search of permanent improvement, there is continuous research in the area in search of new models and algorithms to always provide the most assertive information possible to decision-makers.
As part of this, on April 23, IntellTech participated in an internal seminar at England’s largest distance learning university, The Open University, presenting internal research in the Time Series Forecasting Area.
The Research presented by Intelltech consists of using a modeling method called DBSTAR to predict time series and apply it to instrument readings and compare the predictive performance with WARIMAX, which is a family of statistical models to model time series data, along with GARCH, which assertively predicts how data will change in the future, complementing WARIMAX with the generation of more accurate information.
DBSTAR modeling creates a set of autoregressive Bayesian dynamic models with a smooth transition criterion between them. Currently, this model has been applied to a real series of piezometry, validating the assertiveness of the forecasting generated by this model for real-world time series for later availability to the client in the SHMS.