InfHMM:

Infinit hidden Markov model The code interface is written in R, and for the sake of speed, most parts are written in C++. However, no prerequisite knowledge for both languages is required to run the code. The R file runInfHMM.R sources all needed functions and gives example how to compile and run the code.

This code is used and reviewed in the published paper:

@article{FRIMANE2022331,
title = {Infinite hidden Markov model for short-term solar irradiance forecasting},
journal = {Solar Energy},
volume = {244},
pages = {331-342},
year = {2022},
issn = {0038-092X},
doi = {https://doi.org/10.1016/j.solener.2022.08.041},
url = {https://www.sciencedirect.com/science/article/pii/S0038092X22005916},
author = {Âzeddine Frimane and Joakim Munkhammar and Dennis {van der Meer}},
keywords = {Clear sky index, Probabilistic solar forecasting, Non-parametric Bayesian, -code}
}

Example:

# Clean the memory
rm(list = ls())

# For reproducible results
set.seed(1990)

# You working directory to have all needed source files; e.g.:
setwd("~/")

# needed libs
library(Rcpp)
sourceCpp("FFBScplus.cpp")
sourceCpp("eXpandCplus.cpp")
sourceCpp("countCplus.cpp")
source("InfHMM.R")

### To run the model you need to feed to the infHmm function the following parameters: 
# meas: you data set as vector (time series), 
# n.iter: number of iterations, I recommand 25000, 
# n.burn: number of burning iterations---ignored iterations--- I recommand 10000, 
# ns.init: The starting number of states, I recommande 40 for clear sky index data, 
# pri.gam: the prior gamma, I used c(1,1) in the paper, 
# pri.alf: the prior alpha, I used c(1,1) in the paper.

# load your data, preprocess it and store it in vector.
Your_Data <- runif(500) #loaded.and.preprocessed.data

results <- infHmm(meas = Your_Data, n.iter = 25000, n.burn = 10000, ns.init = 40, pri.gam = c(1,1), pri.alf = c(1,1))

frimane/InfHMM-Infinit-Hidden-Markov-Model

InfHMM:

Example: