Evaluating Effectiveness of Fuzzy Logic as an Interface for Artificial Neural Network

User interface design is a significant part of information system design as user's experience of the system and subsequently the acceptability of the system are based on interface of the system. For effective operation of information system careful design of the user interface is needed. An excellent information system may not be accepted by users if it is not easy to learn and use. Good user interface design can make an information system easy to understand and use, which results in greater user acceptance.;ANNs are powerful tools composed of many simple computational elements interacting across weighted connections. The architecture of ANN is inspired by human brain. ANNs comprise of complicated learning algorithms because of which ANNs have found applicability in a range of problem domains in diverse areas from finance, medicine, engineering , geology and biology. The decision making problems rely on large amount of data and ANNs are suitably applicable for decision making problems as they learn from large amount of data to give the solution. However, ANNs are less convenient to users as they pose problems with interface design. ANNs take as input crisp and normalized data that is very unfriendly with user. The decision making problems involve vagueness, imprecision and uncertainty as a result the input data is in fuzzy form which cannot be directly fed into ANN as it works on crisp data. The decision making problems deals with linguistic parameters whose values are very difficult to mould in equivalent crisp values. Moreover, ANN system lacks explicit explanation and reasoning facility, as knowledge is stored in the interconnections between the neurons. The research work focuses on these limitations of ANN from user interface perspective and emphasizes how fuzzy logic can serve as an effective user interface tool for ANN that will overcome user interface limitations of ANN.;Firstly, it becomes necessary to understand what is an effective interface hence the parameters that make a user interface effective are identified. These parameters are compared against fuzzy logic characteristics to ascertain that fuzzy logic can serve as an effective user interface in general. Having established fuzzy logic as an effective user interface tool the research work is aimed at overcoming user interface limitations of ANN by integrating ANN with fuzzy interface. An experiment is conducted to signify fuzzy logic as an effective user interface for ANN whereby firstly a mere ANN based system is developed to exhibit challenges with user interface design in ANN. A case of decision making problem of employee evaluation is considered for designing and implementing mere ANN based system. The mere ANN based employee evaluation system is enhanced by integrating a fuzzy interface. A hybrid neuro-fuzzy system combines the qualitative approach of fuzzy logic and learning and connectionist capabilities of ANN towards better performance.;The research work encompasses a review process for comparative analysis of ANN based employee evaluation system and neuro-fuzzy employee evaluation system to determine which system has more effective user interface. A review process is conducted among users from diverse fields whereby both the systems were demonstrated to the users and users were asked to rate which system is more effective and satisfactory. The users clearly rated neuro-fuzzy employee evaluation system as better and effective than mere ANN based system.;The research work enhances ANN based employee evaluation system through fuzzy interface where it allows human like perception based approach of inputting evaluation details and also provides reasoning and explanation for the decision being taken resulting into a very user friendly system. Neuro-fuzzy approach can be considered significant methodology for such decision making problems like employee evaluation in real world. Neuro-fuzzy employee evaluation system is a generic product suitable to different companies, which learns from large amount of data and generates employee evaluation output. The system can be used for different other domains like student evaluation, product evaluation, software quality assurance, performance evaluation of a company, website evaluation, with minor modifications.;As far as future scope of the research work is concerned, further research can be done in the area of extracting dynamic domain specific fuzzy membership functions and type-2 membership functions can be defined for employee evaluation parameters. Moreover, each evaluation parameter can be taken as input node to ANN instead of aggregate values of evaluation parameters.;The publications and presentations derived from the research work have been used in various conference and journal papers. In total 8 publications are made out of which 6 are in international journals namely IJESIT (Impact factor: 1.753), IJERMT, IJARCSSE (Impact factor: 2.080), IJEIT (Impact factor: 1.895), JCIES, and IMACST, 1 in international conference-ISSP 2013 and 1 in a symposium on Innovations in Science. (Abstract shortened by ProQuest.).