A Hybrid ANFIS Technique for Effective Performance Evaluation



By Ashwani Kharola

Institute of Technology Management (ITM-DRDO), Mussoorie
Graphic Era University, Dehradun





This study illustrates a hybrid ANFIS based modelling technique for performance evaluation of employees in an organisation. The working model was built in ANFIS Toolbox of Matlab and shows excellent learning ability using hybrid learning algorithm. The proposed model uses a fuzzy rating scale for converting numerical values of attributes into linguistic grades. The ANFIS control strategy provides faster computation with accurate results based on inputs given by leader or experts in an Organisation. The proposed controller is highly effective when the number of input and output attributes is quite large. The simulation results show the validity of the proposed model.

Keywords: ANFIS, Fuzzy, FIS, Performance evaluation, Hybrid, Training, Takagi-Sugeno, Grid Partition, epochs.


Performance evaluation is one of the most important techniques for improving the performance of any individual in an organisation (Arbaiy & Suradi, 2007; Boswell et al. 2002; Fletcher, 2001). It provides a clear performance based feedback to employees (Jawahar, 2006). Performance evaluation involves awarding numerical or linguistic rating grades to employees. It involves judgements which are based on imprecise data when a superior tries to interpret his/her subordinates (Kuvaas, 2006). Fuzzy logic (Hong & Lee, 1996; Hellmann, 2001) is one of the recently developed technique that has created a paradigm shift through many industrial and management applications (Bih, 2006). A lot of research has been done on performance evaluation using fuzzy logic approach. Ingoley and Bakal (2012) proposed a student performance evaluation technique using fuzzy logic. The system takes into consideration of vagueness in question paper besides accuracy rate, complexity and importance. Nunes and O’Neill (2011) described an experiment to evaluate team performance with fuzzy logic reasoning. They implemented a set of Performance evaluation rules which were verified through experimental results. Li and Chen (2009) presented a new method for student learning achievement evaluation by automatically generating the weights of the attributes. The proposed method provides a much fairer and reasonable inference results. Saleh and Kim (2008) proposed a method for student evaluation using fuzzy systems. The proposed system applies fuzzy logic reasoning in considering the difficulty, importance and complexity of solutions. Bai and Chen (2007) presented a method to automatically construct the grade membership functions of lenient, strict and normal type for student evaluation. The system performs fuzzy reasoning to infer the scores of student. Shaout & Yousif (2014) performed a design and implementation of a performance appraisal system using step by step fuzzy inference engine process. The proposed controller was able to select and change parameters such as critical elements, fuzzy method and membership functions.

Yadav and Singh (2014) proposed a fuzzy set and regression analysis based model which is capable of dealing with imprecise and missing data. The model automatically converts crisp set into fuzzy sets by using C-means clustering technique. Singh and Kharola (2013) proposed a fuzzy logic controller for evaluating performance rating of employees. The different attributes were identified and given weights according to relative importance. These attributes were further combined using stage-wise fuzzy reasoning approach. Yadav et al. (2012) described two novel models namely SC-FCM (Subtractive Clustering Fuzzy C-Means) and SC-ANFIS (Subtractive Clustering-Adaptive neuro fuzzy inference system) models for determining students’ academic ranks. These methods not only regulate the division of fuzzy inference system input and output space but also determine the relative member function parameters. In this study a novel Hybrid (Buragohain, 2008) ANFIS controller has been proposed for performance evaluation of an employee in any organisation. The training data for controller has been generated by the knowledge and experience of experts. The designed ANFIS model has three inputs and one output however the number of attributes can be varied. ANFIS provides the flexibility of assigning different weightage to each attribute depending upon their relative importance. The model uses a fuzzy membership rating scale which provides more flexibility in computation as compared to traditional Likert Scale (Edmondson, 2005; Li, 2010). The Simulation results are shown with the help of Matlab-Simulink model which proves the validity of the controller.


To read entire paper (click here)

About the Author


Ashwani Kharola

Dehradun, India


Ashwani Kharola
received B.Tech (Honours) in Mechanical Engineering from Dehradun Institute of Technology (DIT), Dehradun in 2010 and M.Tech in CAD/CAM & Robotics from Graphic Era University (GEU), Dehradun in 2013. He obtained Silver Medal in M.Tech for (2011-13) batch. Currently he is working as Senior Research Fellow (SRF) in Institute of Technology Management (ITM), one of the premier training institutes of the Defence Research & Development Organisation (DRDO), Ministry of Defence, Govt. of India. He is also pursuing PhD in Mechanical Engineering from Graphic Era University. He has published more than 28 National/International papers in peer reviewed ISSN Journals and IEEE Conferences. His current areas of work includes Fuzzy logic reasoning, Adaptive Neuro-fuzzy inference system (ANFIS), Neural Networks, Mathematical Modeling & Simulation, Control of Non-linear systems etc.

Ashwani Kharola can be contacted at [email protected]