Intelligent
Forecaster offers
a variety of functions to analyze and transform time
series. The functionality includes a variety of time
series graphs, a variety of seasonal plots,
autocorrelation plots, and PQ-scatter plots. In
addition to descriptive time series statistics the
functions include automatic tests of stationary
series incl. augmented Dickey-Fuller tests,
seasonality tests, trend-tests, automatic reporting
facilities. It further offers functions to
analyze multiple series ... [find out
more] |
Functionality for cleaning data allows the
imputation of missing values and zero values and to
create dummy time series of outlier values. In
addition explanatory time series of binary or
integer dummies can be created manually or
automatically from iCalendar data. [find out more]
 |
Intelligent
Forecaster offers a variety of forecasting methods,
including neural networks (with various activations
functions, hidden layer dimensions, number of
neurons, and training algorithms), support
vector regression (with various kernel functions and
flexible parameter settings), and statistical
benchmarks of Naive methods, Averages and
Exponential Smoothing (with various forms of
Initializations, Parameter Optimizers, and parameter
ranges) .
[find out more] |
The software allows flexible analysis reports of
multiple error measures (incl. MAE, MSE, MAPE, SMAPE,
MASE, AIC, BIC etc.) for multiple forecasting
horizons across fixed forecasting horizon, rolling
origin evaluation across a set of multiple error
measures using error tables and box plots
[find out more]
.
. |
A
BATCH version of the software Intelligent Forecaster
allows parameter file and command-line
parameter driven, fully automatic BATCH forecasting
of a large number of time series. Performance on
standard INTEL-based Servers and Virtual Servers
allows the prediction of up to 10,000,000
neural network ensembles and time series per hour. [find out more]
 |
|