Results

Training of networks is done by adjusting the neuron weight in the hidden layer to minimize the difference between the two outputs (actual, desired). The network learns by finding a vector of connection weights that minimizes the desired error on the training data set. To conduct this experiment networks are ensemble. At first these networks have five neurons and later on, the number of neurons may be increased up to 40. Among all trained networks the network with the best performance is the most suitable. By repeating the process, it is found that network is best trained by taking 22 neurons in the hidden layer. The accuracy of trained networks at 22 neurons is found to be 99.4%. Results with several different numbers of neurons are shown in Table 5.1.

Case Study

Let’s write an integration assessment for a Patient Data Recording system.

As the name suggests, using this system records various information about patients, such as diseases and their treatment information. Also, this system keeps a record of why particular treatment is given and what in future can be done to treat patients. There must be an option for transferring data digitally. Traditional patient

TABLE 5.1

Training Data Accuracy

Neurons in the Hidden layer

Accuracy

5

89.1%

10

93.1%

12

93.7%

15

96.0%

18

97.7%

20

97.7%

22

99.4%

25

98.9%

30

93.7%

records are more textual. Structural data such as measurement and diagnosis are ingrained in textual data and because of this it is difficult to use for an overview of a patient's condition or statistics. The system should be implemented in such a way that it helps hospital staff to record data in a better manner and can retrieve data in several ways. The major issue is that there is a number of services which provide different information about patients when asked, for example, pathology tests, medicine requirements, ultrasound records, X-ray information, food suggestions, and many more. All of these service providers have their own data format. In computer science we make classes and subclasses, in addition, most of the services are automatically managed through specialized subsystems (production). Some of the tests performed in laboratories are fully automated. Samples with codes are included in the its class, and test findings are obtained in the database shortly afterwards. Databases and the functionality of Data Recording Systems should be integrated with a production system score. These systems do not even have a common communication protocol.

References

Gorton, I.. & Liu, A. (2002). Streamlining the acquisition process for large-scale COTS middleware components. In: J. Dean & A. Gravel (Eds.), International Conference on COTS-Based Software Systems, ICCBSS 2002, LNCS, 2255 (pp. 122-131), vol 2255. Springer, Berlin, Heidelberg

Heiat, A. (2002). Comparison of artificial neural network and regression models for estimating software development effort. Information and Software Technology, 44( 15), 911-922.

Reddy, S., Rao, P. S., Raju, K., & Kumari, V. V. (2008). A new approach for estimating software effort using RBFN network. International Journal of Computer Science and Network Security, 8(7), 237-241.

 
Source
< Prev   CONTENTS   Source   Next >