Add Seasonal Simulation and Multivariate Innovation Simulation Features

.github/dependabot.yml

@@ -0,0 +1,7 @@
+# https://docs.github.com/github/administering-a-repository/configuration-options-for-dependency-updates
+version: 2
+updates:
+  - package-ecosystem: "github-actions"
+    directory: "/" # Location of package manifests
+    schedule:
+      interval: "weekly"

.gitignore

@@ -5,6 +5,11 @@ local_test/
 paper_tests/m4_test/metrics_results/
 paper_tests/m4_test/results_ARIMA/
 paper_tests/m4_test/results_SSL/
+paper_tests/m4_test/init_SSL/
+
 paper_tests/simulation_test/results_simulation_raw/
+paper_tests/mv_probabilistic_forecast/
 
-docs/build
+docs/build
+longhorizon/
+test.jl

Project.toml

@@ -1,7 +1,7 @@
 name = "StateSpaceLearning"
 uuid = "971c4b7c-2c4e-4bac-8525-e842df3cde7b"
 authors = ["andreramosfc <[email protected]>"]
-version = "1.0.2"
+version = "1.1.0"
 
 [deps]
 Distributions = "31c24e10-a181-5473-b8eb-7969acd0382f"

README.md

@@ -153,7 +153,7 @@ using Plots
 airp = CSV.File(StateSpaceLearning.AIR_PASSENGERS) |> DataFrame
 log_air_passengers = log.(airp.passengers)
 
-airpassengers = Float64.(airp.passengers)
+airpassengers = AbstractFloat.(airp.passengers)
 log_air_passengers[60:72] .= NaN
 
 model = StructuralModel(log_air_passengers)

docs/src/manual.md

@@ -149,7 +149,7 @@ using Plots
 airp = CSV.File(StateSpaceLearning.AIR_PASSENGERS) |> DataFrame
 log_air_passengers = log.(airp.passengers)
 
-airpassengers = Float64.(airp.passengers)
+airpassengers = AbstractFloat.(airp.passengers)
 log_air_passengers[60:72] .= NaN
 
 model = StructuralModel(log_air_passengers)

paper_tests/m4_test/evaluate_model.jl

@@ -3,9 +3,9 @@ function evaluate_SSL(
     results_df::DataFrame,
     input::Dict,
     outlier::Bool,
-    α::Float64,
-    H::Int64,
-    sample_size::Int64,
+    α::AbstractFloat,
+    H::Int,
+    sample_size::Int,
     information_criteria::String,
 )
     normalized_y = input["normalized_train"]

paper_tests/m4_test/m4_test.jl

@@ -33,9 +33,9 @@ function run_config(
     results_table::DataFrame,
     outlier::Bool,
     information_criteria::String,
-    α::Float64,
+    α::AbstractFloat,
     save_init::Bool,
-    sample_size::Int64,
+    sample_size::Int,
 )
     NAIVE_sMAPE = 14.427 #M4 Paper
     NAIVE_MASE = 1.063   #M4 Paper
@@ -98,9 +98,9 @@ end
 # Main script
 function main()
     results_table = DataFrame()
-    for outlier in [true, false]
-        for information_criteria in ["aic", "bic"]
-            for α in vcat([0.0], collect(0.1:0.2:0.9), [1.0])
+    for outlier in [true]
+        for information_criteria in ["aic"]
+            for α in [0.1]
                 @info "Running SSL with outlier = $outlier, information_criteria = $information_criteria, α = $α"
                 results_table = run_config(
                     results_table, outlier, information_criteria, α, false, 60
@@ -140,4 +140,4 @@ create_dirs()
 
 main()
 
-run_config(DataFrame(), false, "aic", 0.1, true, 2794)#max sample size
+#run_config(DataFrame(), false, "aic", 0.1, true, 2794)#max sample size

paper_tests/m4_test/metrics.jl

@@ -5,7 +5,7 @@ function sMAPE(y_test::Vector, prediction::Vector)
     return (200 / H) * sum(abs(y_test[i] - prediction[i]) / (denominator[i]) for i in 1:H)
 end
 
-function MASE(y_train::Vector, y_test::Vector, prediction::Vector; m::Int64=12)
+function MASE(y_train::Vector, y_test::Vector, prediction::Vector; m::Int=12)
     T = length(y_train)
     H = length(y_test)
 

paper_tests/m4_test/prepare_data.jl

@@ -1,8 +1,8 @@
-function normalize(y::Vector, max_y::Float64, min_y::Float64)
+function normalize(y::Vector, max_y::AbstractFloat, min_y::AbstractFloat)
     return (y .- min_y) ./ (max_y - min_y)
 end
 
-function de_normalize(y::Vector, max_y::Float64, min_y::Float64)
+function de_normalize(y::Vector, max_y::AbstractFloat, min_y::AbstractFloat)
     return (y .* (max_y - min_y)) .+ min_y
 end
 

paper_tests/simulation_test/evaluate_models.jl

@@ -8,13 +8,13 @@ function get_confusion_matrix(selected, true_features, false_features)
 end
 
 function get_SSL_results(
-    y_train::Vector{Float64},
-    true_features::Vector{Int64},
-    false_features::Vector{Int64},
-    X_train::Matrix{Float64},
+    y_train::Vector{Fl},
+    true_features::Vector{Int},
+    false_features::Vector{Int},
+    X_train::Matrix{Fl},
     inf_criteria::String,
-    true_β::Vector{Float64},
-)
+    true_β::Vector{Fl},
+) where {Fl<:AbstractFloat}
     series_result = nothing
 
     model = StateSpaceLearning.StructuralModel(
@@ -72,13 +72,13 @@ function get_SSL_results(
 end
 
 function get_SS_res_results(
-    y_train::Vector{Float64},
-    true_features::Vector{Int64},
-    false_features::Vector{Int64},
-    X_train::Matrix{Float64},
+    y_train::Vector{Fl},
+    true_features::Vector{Int},
+    false_features::Vector{Int},
+    X_train::Matrix{Fl},
     inf_criteria::String,
-    true_β::Vector{Float64},
-)
+    true_β::Vector{Fl},
+) where {Fl<:AbstractFloat}
     py"""
     import math
     import statsmodels.api as sm
@@ -141,7 +141,9 @@ function get_SS_res_results(
     return series_result, converged
 end
 
-function get_exogenous_ss_inf_criteria(y_train::Vector{Float64}, X_train::Matrix{Float64})
+function get_exogenous_ss_inf_criteria(
+    y_train::Vector{Fl}, X_train::Matrix{Fl}
+) where {Fl<:AbstractFloat}
     py"""
     import math
     import statsmodels.api as sm
@@ -160,13 +162,13 @@ function get_exogenous_ss_inf_criteria(y_train::Vector{Float64}, X_train::Matrix
 end
 
 function get_forward_ss(
-    y_train::Vector{Float64},
-    true_features::Vector{Int64},
-    false_features::Vector{Int64},
-    X_train::Matrix{Float64},
+    y_train::Vector{Fl},
+    true_features::Vector{Int},
+    false_features::Vector{Int},
+    X_train::Matrix{Fl},
     inf_criteria::String,
-    true_β::Vector{Float64},
-)
+    true_β::Vector{Fl},
+) where {Fl<:AbstractFloat}
     best_inf_crit = Inf
     current_inf_crit = 0
     coefs = nothing

paper_tests/simulation_test/simulation.jl

@@ -327,7 +327,7 @@ CSV.write("paper_tests/simulation_test/results_metrics/metrics_confusion_matrix.
 
 df_mse_bias = DataFrame()
 
-function convert_to_sci_notation(num::Float64)
+function convert_to_sci_notation(num::AbstractFloat)
     # Get the exponent part of the number in scientific notation
     exp_part = floor(log10(abs(num)))
 

paper_tests/simulation_test/simulation_generator.jl

@@ -23,7 +23,7 @@ function assert_invertibility(q::Vector{Fl}) where {Fl}
     return all(abs.(roots_of_inverse_polinomial(poly)) .< 1)
 end
 
-function generate_sarima_exog(T::Int64, M::Int64)
+function generate_sarima_exog(T::Int, M::Int)
     X = zeros(T, M)
     s = 12
     for j in 1:M
@@ -87,7 +87,7 @@ function generate_sarima_exog(T::Int64, M::Int64)
     return X
 end
 
-function generate_subset(T::Int64, M::Int64, K::Int64)
+function generate_subset(T::Int, M::Int, K::Int)
     s = 12
 
     μ1 = 1.0