Identifying Relationships Between Eco-innovation and Product Success - Technology Management for Sustainable Production and Logistics - page 183

Table 5 Comparison of
RMSE for different models

Model RMSE

Learning set Testing set

2.0106 3.2662

ANN—GDX

ANN—LM 0.00003 4.4569

Linear model 0.9620 15.5144

but the RMSE for the testing set was greater than that obtained through the use of
the GDX algorithm, which can result from the overtraining of ANN and the lack
of its generalization abilities. It is noteworthy that RMSE generated with the use
of ANNs is smaller for testing set than RMSE for the linear model. In the next
step, the ANN structure can be used to forecasting of the dependant variable (net
profit for a product that is in the development process).

5.3 Forecasting and If-What Analysis

The non-linear relationships between input and output data are stored in the struc-
ture of a neural network. If new data for a product in the development process is
lead to the trained network, then the result is the forecast of net profit for a prod-
uct. Let us assume that for the actual project data, 35 variables are input to ANN.
As a result, the output of ANN indicates the projected net profit for a product.
Moreover, the use of what-if analysis can indicate the directions of changes in pro-
ject environment that should increase the probability of net profit for the devel-
oped product (Relich 2014 ).
Table 6 presents a two-dimensional analysis for number of ideas and number
of team members in a new product development. For instance, if number of team
members equals 4 and number of ideas generated in the development process of
product equals 40, the forecast of net profit equals 4.84 monetary units. The deci-
sion-maker can use this analysis for simulating net profit value for various criteria.
The presented analysis can be extended to support a higher number of dimen-
sions (for others input variables) and towards the sensitivity analysis to support the
decision-maker in the choice of project environment parameters that can be changed.

Table 6 Example of what-if
analysis

Number of ideas for
developing a new product

Number of team
members for developing
a new produ ct
4 5

3.76 3.90

30

3.94 4.35

35

40 4.84 5.14

7.82 9.42

45

11.32 12.41

50