Factors Affecting Component Integration Efforts
Components are pure black box entities. No-one can check the source code of components. Interaction with these components is possible only through interfaces. Hence, the interface acts as the main source to understand, implement, maintain, and reuse components. These interfaces explain individual elements of a component systematically.
Hence, complexity of these components is very crucial for correctly estimating overall component complexity.
The understanding of a component may be defined as the familiarity of the component with the environment. The environment can be both software and hardware. So, if a component is very familiar with the environment then integration efforts will be fewer, otherwise it will be high.
By Component Quality we mean that we select a well-suited component for our application. If the selection of a component is right, then the integration efforts will be low.
Artificial Neural Network-Based Approach
In this chapter we have discussed the Artificial Neural Network (ANN) technique, measuring the integration efforts of a software component. ANNs are impressive techniques in the area of clustering and classification [MAINT l, MAINT 2]. The following are the reasons for selecting Neural Networks:
- • An ANN is adaptive and can classify patterns easily.
- • The complexity of a network is adjusted by an ANN to that of a problem, hence, it produces better outcomes as compared to other analytical models.
- • Neural Network techniques are known as “black boxes”. Users cannot understand the process easily.
- • There are no fixed guidelines for considering neural networks.
Neural Network Architecture
Figure 5.1 describes an elementary neuron with n input. A weight value w is assigned to each input. Transfer function/gets its input from the sum of all weighted inputs. For output generation, neurons may use any differentiable function.