raw2encoded: Cluster meter data into the best fit shape clusters and...

In ConvergenceDA/visdomloadshape: R package for load shape clustering and analysis, extending the visdom package

Description

Useful for generation of a load shape cluster center dictionary and encoding in one shot

Usage

raw2encoded(rawdata, is.clean = T, use.all = T, s.size = 1e+05,
  target.size = 1000, mode = 1, d.metric = 1, ths = 0.2,
  iter.max = 100, nstart = 1, two.step.compress = F, verbose = F)

Arguments

is.clean	whether to do data interpolation or not. Note that the default interpolation method is very memory and CPU intensive. You may be better off doing your own interpolation or passing only complete.cases().
use.all	whether to use all data to generate a dictionary or not
s.size	sample size to use to generate a dictionary
target.size	target size of the dictionary (i.e. nubmer of clusters)
mode	1: use ths1, 2: use ths2, 3: use ths3, 4: use ths4
d.metric	1: use euclidean distance metric, otherwise use cosine distance metric
ths	will be transferred to akmeans parameter according to mode setting ths1: threshold to decide whether to increase k or not: check sum((sample-assigned center)^2) < ths1*sum(assigned center^2) ths2: threshold to decide whether to increase k or not: check all components of |sample-assigned center| < ths2 ths3: threshold to decide whether to increase k or not: check inner product of (sample,assigned center) > ths3 , this is only for cosine distance metric ths4: threshold to decide whether to increase k or not: check sum(abs(sample-assigned center)) < ths4
iter.max	maximum iteration setting to be used in kmeans
two.step.compress	whether to reduce the dictionary only by hierarchical clustering or hier+use top N shapes. this option gets activated only when the ratio (original dictionary size before compression/target.size) is larger than 10
verbose	whether to show log or not
rdata	rawdata of format n by p matrix
n.start	parameter to be transferred to kmeans

ConvergenceDA/visdomloadshape documentation built on May 8, 2019, 8:34 a.m.