At ISAGEN S.A., I developed regression and time series forecasting models in Python using geotechnical instrumentation data. The models helped identify behavioral patterns, detect potential anomalies, and support operational decisions for the Dam Safety team.
I also designed a method for setting multivariate statistical thresholds by clustering instrumentation data according to temporal behavior. This improved anomaly detection and reduced false alarms in instrument monitoring.
To make this work accessible to the team, I developed and deployed an interactive Streamlit application that automates the Hydrostatic-Seasonal-Time (HST) model and threshold generation, so team members can analyze new instrumentation data independently.
At the 33rd International Symposium on Statistics, I presented forecasts of Colombia’s 2024 quarterly mining and quarrying GDP using classical time series and Bayesian state space models. I also developed two R Shiny applications for visualizing failure and censored data, supporting model selection, goodness-of-fit assessment, and reliability analysis.
