write a code to identify states in timeseries data with hmm in r

main.r
library(HMM)

# create a simulated timeseries data with 3 states
set.seed(123)
n <- 100
x <- c(rnorm(n/3, mean=0, sd=1), rnorm(n/3, mean=2, sd=1), rnorm(n/3, mean=-2, sd=1))

# define number of states
n_states <- 3

# build HMM model
hmm_model <- initHMM(c("state1", "state2", "state3"), c("A", "B"), transProbs=matrix(c(.6, .2, .2, .2, .6, .2, .2, .2, .6), n_states, n_states))

# normalize the observation data
norm_x <- (x - mean(x))/sd(x)

# fit HMM model to normalized data
fit <- BaumWelch(norm_x, hmm_model$hmm)

# Viterbi algorithm to identify states
viterbi_path <- viterbi(norm_x, hmm_model$hmm)

# plot original data with identified states
plot(x, type="l", main="HMM State Identification")
for (i in 1:(length(viterbi_path)-1)) {
  lines(c(i, i+1), c(x[i], x[i+1]), col=viterbi_path[i])
}
801 chars
28 lines