| TRMF_seasonal | R Documentation |
Creates a regularization scheme that favors seasonally varying solutions and adds it to a TRMF object. In matrix optimization form, it adds the following term to the TRMF cost function: R(x) = lambdaD^2||w(DX_s)||^2 + lambdaA^2||X_s||^2 where X_s is sub-set of the Xm matrix controlled by this model and D is a (with a lag of freq) finite difference matrix.
TRMF_seasonal(obj,numTS = 1,freq = 12,sumFirst=FALSE,lambdaD=1,lambdaA=0.0001,weight=1)
obj |
A TRMF object |
numTS |
number of latent time series in this model |
lambdaD |
regularization parameter for temporal constraint matrix |
lambdaA |
regularization parameter to apply simple L2 regularization to this time series model |
weight |
optional vector of weights to weight constraints, i.e. R(x) = lambdaD^2*||w*(D%*%X)||^2 |
freq |
The frequency of the seasonal time series model. Minimize the differences of lag = |
sumFirst |
minimize the sum of first |
TRMF_seasonal(freq=N) fits a lag N random walk. For monthly data, use freq=12, for quarterly data, freq=4. If sumFirst = TRUE, the sum of the first freq elements in the latent time series are also minimized.This can be used to help force the seasonal component to vary around a zero mean.
Returns an updated object of class TRMF.
Chad Hammerquist
Yu, Hsiang-Fu, Nikhil Rao, and Inderjit S. Dhillon. "High-dimensional time series prediction with missing values." arXiv preprint arXiv:1509.08333 (2015).
create_TRMF, TRMF_columns,TRMF_simple, TRMF_trend
# create test data tm = 3*poly(x = (-20:20)/10,degree=3) sm = diffinv(rnorm(29,0,.1),lag=12,xi=(-5:6)/6) xm = cbind(sm,tm) fm = matrix(runif(40),4,10) Am = xm%*%fm+rnorm(410,0,.1) # create model obj = create_TRMF(Am) obj = TRMF_columns(obj,reg_type ="interval") obj = TRMF_trend(obj,numTS=3,order=2) obj = TRMF_seasonal(obj,numTS=1,freq=12,lambdaD=5) out = train(obj) plot(out)