multinomDirClass: Nonparametric Multinomial Mixture Modeling

In frimane/SolMultinomClass: Solar Radiation Classification Using Nonparametric Multinomial Mixture Modeling

Description

Nonparametric Dirichlet-Multinomial classification to rows of a data matrix without specifying the number of classes (nonparametric), is also a random variable.

Usage

multinomDirClass(measures, n.it = 10000, n.B.in = 3000, zi = rep(1, N),
alf = 3, seed = 2308.2202, talk = T)

Arguments

measures	matrix(), of data, each row as multinomial vector, can be the output of the function codinData()
n.it	numeric(1), number of iterations
n.B.in	numeric(1), number of burn-in iterations (ignored iterations)
zi	vector(), initialization of indicator variable.
alf	numeric(1), concentration parameter
seed	numeric(1), reproducibility of the results (same results with the same seed)
talk	logical, shows the classification evolution in real time, if talk is true.

Value

An object of class list, with elements:

likely class sequence	vector of the class indicators sequence
likely parameters	list of parameters of each class
likely hyperpameter	parameter of the hyperprior
likely concentration parameter	most likely concentration parameter
Sampling sequence of concentration parameters	if you want to see the posterior alfa distribution
sequence of class numbers	if you want to see the posterior class number distribution
number of elements in the likely class	counts at each combination of factor levels within likely class sequence
calculation time	the duration of calculates

Author(s)

Azeddine Frimane.
Laboratory of renewable Energies and Environment (LR2E),
Faculty of Science, IBN TOFAIL University, Morocco.
email: [e-mail]; [e-mail]

Examples

# load needed library
library("SolMultinomClass")

# load data
data("OregonUData") 

# calculate the extraterrestrial radiation for the given site
mat2 <- rayExt(phi = 43.12, lg = -121.06, tStep = 300) # tStep = 5 minutes

#In accordance with considered data, 12 hours of measurements by 5 min, from 6h to 18h.
mat2 <- mat2[,73:216] # 

# now, coding the data as multinomial distribution of the clearness index (mat2/mat1)
mat <- codingData(OregonUData, mat2, 8) # 8 bins

# finally, carry out the classification, number of iterations must be large (15000 only fo example)
classification <- multinomDirClass(measures = mat, n.it = 15000, n.B.in = 3000)

# save the classification results to a file
save(classification, file = "classificationResults.RData")

frimane/SolMultinomClass documentation built on Jan. 18, 2021, 4:49 a.m.