Time series modeling via JuMP

This talk explores the application of JuMP, a powerful optimization modeling language, to time series modeling. Traditional statistical packages often underutilize modern optimization capabilities and the potential of existing solvers when addressing the optimization aspects of time series model estimation and specification. We delve into three distinct models-SARIMAX.jl, StructuralScoreDrivenModels.jl, and StateSpaceLearning.jl-highlighting the optimization prowess they bring to SARIMA, Score Driven Models, and State Space Models, respectively. In addition, we substantiate the efficacy of JuMP-based time series modeling through numerical experiments, demonstrating notable improvements in performance over traditional benchmarks.