module and stored in the knowledge base (KB). The
KB is built on the ontology of innovation activity and
economic potential. It also requests data to be
displayed in the web interface by the semantic search
system for subsets of data.
There are several types of classifications of
forecasting methods. According to one of them, all
forecasting methods can be divided into heuristic
methods, which are based on the predominance of
intuition, i.e. subjective principles, and economic and
mathematical methods, which are dominated by
objective principles (Sonina, 2014). To get a
completer and more objective picture when building
an empirical model for predicting innovation activity,
we will consider methods from both groups. The
vector autoregressive model and dynamic stochastic
model of general equilibrium are chosen as an
example of economic and mathematical approaches,
while the technological foresight method represents
heuristic methods.
Thus, several competing methods are used in the
decision-making module:
* Vector autoregressive model (VAR)
* Dynamic stochastic general equilibrium model
(DSGE)
• Neural network.
The module works as follows. The user of the DSS
submits the necessary parameters for input by
selecting them from the KB:
- For a neural network, the user selects a set of
economic indicators for the past years, based on
which it will be trained.
- For the vector autoregressive model, the user
chooses indicators for forecasting the innovation
activity of the economy as variables for building the
VAR model and the period on which the forecast will
be made.
- For DSGE, a list of parameters are selected. Some
of them are set as constants, taking into account
adaptation for the Russian economy.
This is followed by a prediction process, and
results that can be installed by the user in the decision
method adjustment module (for example, by
changing the values of input parameters or adding
them). The results obtained are used by experts in
determining forecast data, and can also be introduced
in the foresight system.
Despite the advantages of the DSGE model, there
is a criticism of it, which highlights the following
shortcomings (Andrianov et al., 2014):
* Using the concept of rational expectations, which
assumes that economic agents make the most
effective use of all available information and all
available experience when making decisions;
* Using the representative agent principle, which
reduces complex economic systems to separate
elements, and as a result neglects holistic basis of a
system;
* Applying filters depending on the selected method,
smoothing parameters, initial and final filtering
periods, and so on.
* Time-consuming procedure for deriving and
parameterizing equations.
Therefore, in the course of the study, it seems
appropriate to search for an analogue for this method,
which most fully meets the tasks set.
3.2 UML Diagram of Options for using
the Decision Methods Module
The UML diagram of the options for using the
decision-making methods module is shown in Fig.2.
Firstly, it is necessary to select the type of forecast
model.
Аlso, economic indicators and forecast period are
chosen. The indicators are submitted to the model as
input data. In the DSS, these indicators are stored in
the previously created ontology of macroeconomic
and statistical data.
Next, preliminary steps before forecasting are
performed. For DSGE models it is the calibration of
input parameters, for the VAR model – definition of
the maximum length of the lag, for the Neural
network - training on selected economic indicators
during the particular time period.
After that, the user receives the result of the
selected forecasting models, which can be viewed in
tabular and graphical form, filtered by different
criteria. The results obtained by different methods
can be compared with each other.
The models are built iteratively: each run can be
followed by a process of adjusting the results, which
changes the input parameters, and for the neural
network – training on the adjusted data.
Thus, the decision-making module works as a
"black box" for the end user: the user selects
economic indicators to submit for input (the
indicators themselves and their values for N years are
stored in the ontology), and at the output receives the
result of forecasting indicators.
3.3 UML-diagram of the Class of the
Decision Methods Module
UML-diagram of the class of the decision methods
module is shown in Fig. 3. Let us look at it