See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/358702813

Soft Computing Approaches with Neural Networks -An Evaluation

Article in Journal of Advanced Research in Dynamical and Control Systems · November 2018

CITATION READS
1 79

2 authors, including:

M. Jeyakarthic
Annamalai University
83 PUBLICATIONS 217 CITATIONS

SEE PROFILE

All content following this page was uploaded by M. Jeyakarthic on 20 December 2023.

The user has requested enhancement of the downloaded file.

See discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/358702813

Soft Computing Approaches with Neural Networks -An Evaluation

Article in Journal of Advanced Research in Dynamical and Control Systems · November 2018

CITATION READS
1 63

2 authors, including:

M. Jeyakarthic
Annamalai University
76 PUBLICATIONS 187 CITATIONS

SEE PROFILE

All content following this page was uploaded by M. Jeyakarthic on 18 February 2022.

The user has requested enhancement of the downloaded file.

Jour of Adv Research in Dynamical & Control Systems, Vol. 10, 14-Special Issue, 2018

Soft Computing Approaches with Neural

Networks – An Evaluation
M. Jeyakarthic, Dept. of Computer and Information Science, Annamalai University, Annamalai Nagar.
E-mail:[email protected]
N. Subalakshmi, Computer Science and Engineering Wing, Annamalai University, Annamalai Nagar.
E-mail:[email protected]
Abstract--- Neural Networks (NN) is a term applied to a network of neurons, similar to biological neuron networks.
It is the base for mat such networks including Artificial Neural Networks (ANN), where ANNs are interconnected
artificial neurons. ANNs mimic the properties of biological neurons. In machine learning, NNs can take finite
number of inputs for predicting outputs or outcomes. These networks have the ability improve generic
performances, making them an important factor of Soft computing (SC), where algorithms learn using reinforced
learning. It aims to find solutions for imprecise or uncertain data which otherwise may not produce any results. It
integrates soft computing techniques and tools. It also yields rich knowledge and flexible inferences. It
accommodates acquisitions and supple knowledge by enabling systems like Intelligent Information Systems (IIS) or
Artificial Intelligence (AI) to be built at a very low cost. This paper highlights the importance of Neural networks in
soft computing with a model application.
Keywords--- Soft Computing, Neural Network, Artificial Intelligence, Teaching Method Evaluations.

I. Introduction
The term Soft Computing applies to multiple computer disciplines. SC fuses methodologies in its models for
solving real-time problems which do not have proper mathematical solutions [1] and is considered a part of artificial
intelligence. The validity of even approximate solutions by using mathematical tools to such complex problems in
soft computing is supported, where hard computing and soft computing are technologies for solving such complex
problems. [2]. Hard Computing creates fast and accurate solutions for precise information, while soft computing
approximates complex problems [3]. Solutions for near impossible problems are possible in soft computing as they
do not tie themselves to precisions. Soft computing problems are frequently modeled on vague boundaries like 0 to 1
[4], churning solutions. These problems while using conventional computations work only on specified value
ranges. Though SC became a formal field of study in early 1990s, only recently, interest in this area has grown
rapidly. Older approaches could not find solution to complex problems in the fields of management science, biology,
humanities, medicine, [5]. They often remained stubborn to solutions through mathematical and analytical methods.
This made SC an emergent methodology for achieving robust solutions for complex problems at very low costs. SC
combines methodologies (fusion) for computational results. There are three main branches in soft computing namely
evolutionary computation, fuzzy Systems and artificial neural computing [6]. SC is a field of mathematical and a
computer study which has the ability to solve real world problems modeling the human brain, unlike traditional
computing models. SC techniques are liberal in accepting half truths, imprecision and uncertainties in data and solve
them by approximations.

Fig. 1. SC Problem solving Approch

ISSN 1943-023X 1001

Received: 20 October 2018/Accepted: 15 November 2018
Jour of Adv Research in Dynamical & Control Systems, Vol. 10, 14-Special Issue, 2018

Even complex computational problems can be offered workable solutions using soft computing approximations
where techniques complement each other. Hard computing models complete truths in solutions, leaving gaps in
computing of imprecise and uncertain data. Inductive reasoning plays a larger role in SC. This paper highlights SC
problem solving abilities with the application of NNs and can be of help to other studies. Figure 1 depicts SC
Problem solving Approach.

II. Applications of Soft Computing

Global data has grown tremendously with the arrival of computer and electronic devices [7] which can be
attributed to several factors like internet, sensors, electronic device integrations and the search for human dependent
information [8]. Complex data structures make it difficult for traditional computer processing applications [9]. For
example, the banking sector is experiencing voluminous data hurdling customary processes, due to the increase in
recorded banking data, [10]. Moreover, the threats to security have also added to this rise in banking information as
detailed information is trapped [11]. On the healthcare sector, medical centers and units are recording heavy increase
in data by integrating ICT and healthcare service information. This integration has provided the need for faster
information processing to facilitate prompt diagnoses and treatment. Huge investments are needed for creating
analytical tools to manage this data [12]. Though analytic tools needed for the educational sector can solved by
simulations or validations, technical experts have been exposing required tools, due to deficiency of required tools
and conventional software, creating an issue [13]. Complex problem solving of imprecise or uncertain data can find
solutions in soft computing as they determine an object’s proximity based on partial truths and approximations in
boundaries for achieving practicability and robustness. SC is formed on machine learning techniques helping learn
without including programming [14]. Activities and applications in soft computing have been increasing steadily
with focus on solutions for complex real-world problems. Soft computing is also particularly useful when
applications are stranded with insufficient information or complex data. The acceptance of soft computing to
approximations instead of absoluteness, allows researchers and learners to approach problems which cannot be
solved by traditional computing processes. In mathematics, SC can be used for redesigning even in well-established
areas like differential equations to deal with vagueness. Soft computing has found use multimedia applications as
they have strong cognitive and learning abilities. Soft computed multimedia uses include image processing and
biometric applications Uncertain economic environments lead to vague and uncertain inputs, where soft computing
can play a major role in getting conclusions. Soft computing economics and business can be extended to analysis of
customer trends and attitudes, in finance like stock market predictions, portfolio management, loan assessments and
management of risks. In electronic business, it can be applied to risk analysis in e-commerce. Further, the internet
with its voluminous information is also a good candidate for soft computing. It can be used in ontology and semantic
based analysis of probabilistic information and more specifically in web data researches. Thus, systems can
accommodate abstract data based on surrounding information and taught to react correspondingly using NNs, a part
of SC. Application of NNs have shown its application in fault diagnosis (Dynamic and Static) and industrial process
controls in process industries. Soft computing techniques have found wide applications, due to the above detailed
reasons.

III. Soft Computing and Neural Networks

Soft computing techniques bear a resemblance to human reasoning in sharp contrast to traditional techniques
which are largely based on conventional logic or mathematical capabilities of a system [15]. Similarly, NNs are an
emulation of human biological neural systems. Neural Networks introduced by Zadeh [16], can translate qualitative
knowledge for solving problems complex problems. NN’s interpolative reasoning is its main characteristic. They
also work in a distributed or parallel manner by composing interconnected processing elements. These processing
elements adapt by changing concurrently to information flow rules. Moreover, even if an element fails, other
elements continue in their paths reducing problems, mainly due to the parallel nature of NNs while processing [17].
Neural network’s capability to learn on their own reduces reprogramming needs and finds use in many applications.
Thus, a neural network can perform tasks that linear programming cannot perform.

IV. Soft Computing Teacher Quality

Higher educational institutes need quality assurance while imparting education which can be achieved by
providing authentic evaluations of teaching qualities. Further, the quality of teaching is dependent on several factors
which need to be measured for achieving defined quality outcomes.

ISSN 1943-023X 1002

Received: 20 October 2018/Accepted: 15 November 2018
Jour of Adv Research in Dynamical & Control Systems, Vol. 10, 14-Special Issue, 2018

V. Methodology
Soft computing models can be achieved using neural networks. This paper uses NNs to examine teaching quality
of with statements from which accurate inferences are difficult to be drawn. Subjective attributes collected from
students of different colleges in and around Chidambaram are fed to a neural network. The lists of institutions were
accessed from websites, personal contacts. Subjective questions comprised of perceptions related to teaching
quality. Three dimensions namely learning outcomes, Responsiveness, and Academics were considered in the
questionnaire as listed in Table I.
TABLE I. Learning Dimensions

NN’s neuron weights are initialized randomly and later updated for optimizing results using Back-propagation
technique. This works used back propagation to optimize weights, due to its efficiency in achieving gradients
efficiently while minimizing the loss in a NN. Obtained responses from back-propagation algorithm’s outputs from
the training and test data of the survey, were used to assess teaching quality owing to NN’s generalization ability.
The network consisted of three layers namely input (In), hidden layer (Hn) and a single node output layer. The
overall teacher service quality was the output. Hidden layer (Hn) nodes was decided by (1)

(1)
Where, Hn= Normalization of raw data for getting values in a common scale between 0 and 1. The learning rate
of the training phase was set a value lesser than 0.1 with momentum parameter near to 0. Seventy-five percent of
stakeholder data was used in training, while the rest was put to test. Perception-only model was used to predict
service quality and only perceptions based inputs was fed to the NN and in agreement to the guidelines pertaining to
defined service-quality measurements. The perception of students is more significant than the perceptions and
expectations gaps. Table II lists the Training and testing Dataset details of the Perceptions-only Model..
TABLE II. Perceptions-Only Model Training and testing Dataset
Groups Size1 Size2 Size3
Training 75% 75% 75%
Testing 15% 18% 25%

VI. Results and Discussions

This section discusses the experimental results obtained in the work. Figure 2 depicts the neural network adopted
on the data set for the input and output features.

ISSN 1943-023X 1003

Received: 20 October 2018/Accepted: 15 November 2018
Jour of Adv Research in Dynamical & Control Systems, Vol. 10, 14-Special Issue, 2018

NN Inputs and Outputs

Fig. 2.
The experiments achieved different results in training and testing stages. By using these results, the performance
teaching quality was analyzed. The performance of the output was analyzed based on different factors like accuracy,
estimation time, and prediction error, where accuracy was calculated using (2).
Accuracy=(TN+TP)/(TN+FP)+(TP+FN) (2)
Where, TP is a True Positive; FP is a False Positive; TN is a True Negative; FN is a False Negative. The
training and testing results obtained is listed in Table III and depicted in Figure 3.
TABLE III. Perception-Only Performance in Training and Testing
Training Training Training Error Testing Size Testing Accuracy Testing Error
Size Accuracy

75% 96.28% 3.72% 10% 94.24% 5.76%

75% 94.215% 5.79% 18% 89.00% 11.00%
75% 95.324% 4.68% 25% 95.00% 5.00%

Perception Only Model Performance

Fig. 3.
It is evident from Figure 3 that the performance of the model does not differ much between training and testing.
On size 3 they are almost similar in terms of error rates with accuracy, thus validating the model. High accuracy
obtained in the size 75% training with 96.283%, and 10% testing with 5.76% errors. On the other hand, the lowest
performance accuracy results achieved in training was 94.215%.

VII. Conclusion
This paper investigated and evaluated teaching quality outcomes based on vague descriptions of perceptions. The
quality of teaching was emphasized using Neural Network algorithms On deep examinations, about 90% prediction
accuracy was achieved. It can be concluded that NNs and soft computing is an excellent resource for complex
problems and problems where there are subjective and vague data from which conclusions are difficult to be
inferred.

ISSN 1943-023X 1004

Received: 20 October 2018/Accepted: 15 November 2018
 Jour of Adv Research in Dynamical & Control Systems, Vol. 10, 14-Special Issue, 2018

References
[1] Chaturvedi, D.K. Soft computing: techniques and its applications in electrical engineering, Springer, 2008.
[2] Gupta, P. and Kulkarni, N. An Introduction of Soft Computing Approach over Hard
Computing. International Journal of Latest Trends in Engineering and Technology (IJLTET) 3 (1) (2013)
245-258.
[3] Ibrahim, D. An overview of soft computing. Procedia Computer Science 102 (2016) 34-38.
[4] Zadeh, L.A. Fuzzy logic, neural networks, and soft computing. Communications of the ACM 37 (1994)
77-84.
[5] Kaczmarczyk, K., Wit, A., Krawczyk, M., Zaborski, J. and Piłsudski, J. Artificial Neural Networks (ANN)
applied for gait classification and physiotherapy monitoring in post stroke patients. In Artificial Neural
Networks-Methodological Advances and Biomedical Applications. InTech, 2011.
[6] Bonissone, P.P. Soft computing: the convergence of emerging reasoning technologies. Soft computing 1 (1)
(1997) 6-18.
[7] Boyd, D. and Crawford, K. Six provocations for big data. Social science research network: A decade in
internet time: Symposium on the dynamics of the internet and society. Elsevier, Netherlands, 2011.
[8] Segaran, T. and Hammerbacher, J. Beautiful data: the stories behind elegant data solutions. O'Reilly
Media, Inc, 2009.
[9] Dedić, N. and Stanier, C. Towards differentiating business intelligence, big data, data analytics and
knowledge discovery. International Conference on Enterprise Resource Planning Systems, 2016, 114-122.
[10] Ibrahim, T.H., Abaker, Y., Ibrar, B.A., Nor, M. and Salimah, G. Big data on cloud computing: Review and
open research issues. Inf Syst., 47 (2015) 98-115.
[11] Wu, D., Rosen, D.W., Wang, L. and Schaefer, D. Cloud-based design and manufacturing: A new paradigm
in digital manufacturing and design innovation. Computer-Aided Design 59 (2015) 1-14.
[12] Lee, J., Lapira, E., Bagheri, B. and Kao, H.A. Recent advances and trends in predictive manufacturing
systems in big data environment. Manufacturing Letters 1 (1) (2013) 38-41.
[13] Preis, T., Moat, H.S., Stanley, H.E. and Bishop, S.R. Quantifying the advantage of looking
forward. Scientific reports 2 (2012).
[14] Wernick, M.N., Yang, Y., Brankov, J.G., Yourganov, G. and Strother, S.C. Machine learning in medical
imaging. IEEE signal processing magazine 27 (4) (2010) 25-38.
[15] Katkar, G.S. The Assembly of Neural Network in Soft Computing. International Journal of Advanced
Research in Computer Engineering & Technology (IJARCET) 1 (3) (2012).
[16] Soni, M.A., Mehta, M.S. and Gaurav, A.R. The convergence of budding reckoning Soft computing
technologies. International Journal on Recent and Innovation Trends in Computing and Communication 1
(4) (2013) 385-387.
[17] Shah, B. and Trivedi, B.H. Artificial neural network based intrusion detection system: A
survey. International Journal of Computer Applications 39 (6) (2012) 13-18.

ISSN 1943-023X 1005

Received: 20 October 2018/Accepted: 15 November 2018

View publication stats