International Academic Institute for Science and Technology International Academic Journal of Innovative Research Vol. 3, No. 2, 2016, pp. 1-10. ISSN 2454-390X International Academic Journal of Innovative Research www.iaiest.com Applications of Fuzzy Logic and Artificial Neural Networks in Evaluation and Ranking of Teachers Based on Framework for Teaching Model Sara Abedi Kooshki 1, Hassanreza Zeinabadi 2, Ahad Jafarnezhad 3,Hadi Abedi kooshki 4 1 MA student, Department of Educational administration Faculty of Management Kharazmi University, Iran Corresponding author 2 Associate professor, Department of Educational administration Faculty of Management Kharazmi University, Iran 3 MA AI, Ministry of Education, Tehran, Iran 4 Department of Engineer Semnan University, Iran Abstract Evaluation of teachers is considered as one of main educational policy. And suggests teaching improvement strategies to achieve educational development. Framework for teaching is one of evaluation models to achieve this goal (Danielson,1996). This model includes four, planning and preparation, the classroom environment,. instruction, and professional responsibilities aspects. These aspects describe properties and behaviors of teachers. According to ranking, teachers classified in four groups by their performance. This method based on four-valued logic. Evaluation of teachers by four values leads to failure in relative justice, unrealistic equality and inequalities. Also, there are complexities of evaluation process and probable human errors in this method. In this study, we will develop a solution based on Neuro-fuzzy evaluation system. In this method the inputs of the system are based on fuzzy logic that includes advantage of every criterion. Outputs of fuzzy will be inputs of the neural network system and finally we will isolate patterns by artificial neural network. This system will be implement in the software environment and it makes the evaluation process easier and more precisely than before. Overall, this model could be as a basement method for precise evaluation of educational activities, deserve to achieve educational degrees, ranking and payment of teachers salaries in education system. Keywords: Educational Evaluation, Fuzzy System, Artificial Neural Network, Teacher. 1
Introduction: Evaluation of teacher is one of main educational policies and a strong strategy to improve educational development of school and university students. To increase our awareness from responsibility and ability of teachers, evaluation of teachers by researchers considered as a solution improve teachers efficiency, a tool to judge about them and tell of deficient teachers(avalosa & Assael, 2006 ). It is a main factor in paying salaries and benefits and helps managers to decide about teachers performance (Kimball & Milanowski,2009), improve teachers efficiency and support their professional development (danielson, 2001; Stronge & Tucker, 2003; quoted from Tuytens & Devos, 2011). Considering the importance of the evaluation of teachers, the main goal of this study is description of different views about evaluation and identifying a software model, by using artificial intelligence techniques for that. One main reason in lack of a consensus about evaluation concept is uncertainty in measurement methods and performance measuring. In this study, by considering uncertainties in concepts and performance measuring methods we will use verbal phrases and fuzzy logic to become certain about uncertain matters. To decrease mind interferences in classification process, we used artificial intelligence neural network and it is a strong techniques in artificial intelligence. Framework for Teaching By reviewing properties of effective teachers among theoretical and practical researchers, Danielson (1996) proposed a framework for teaching model to improve and raise learning of students. Development of this framework is based on constructivist learning theory and it was designed by national professional teaching standards board to consider knowledge and skills of teachers (Mangiante, 2011). This model includes four planning, preparation, class environment and professional responsibility aspects. These aspects have 22 dimensions and there are 2-5 components for each dimension. At all, dimensions of the model have 66 components. These dimensions describe properties and behaviors of teachers. To analyze the results of the model and based on their ranks in four aspects, teachers classified in four competency performance groups. Orders of level of performance are trainee, amateur, professional and advanced. Researches based on components of this model, indicate that teachers with higher score have students with acceptable educational status (Stronge et al, 2011). Fuzzy Logic Fuzzy logic is based on fuzzy set theory tries to follow human methods in presenting and reasoning in real world when encounter with uncertainty. Uncertainty may be caused by popularity, chance, ambiguity, weakness or lack of knowledge. In mathematics fuzzy set is described by attribution of values to every possible matters in reference set. These values shows membership degree in fuzzy set. Otherwise, fuzzy sets provide flexible membership degree for every member of the set. With this description, fuzzy logic could be a strong tool to solve different managerial problems (Shaverdi et al, 2013). Artificial Neural Network Neural network is a set of many connected processor elements known as neurons. This network inspired from brain neural network. Neural networks are capable to learn from examples. We could train them by presenting well known examples of a matter and this will increase their knowledge about it. If training is properly done, the network will be able to solve unknown and not-trained problems. In this study, we will train the artificial neural network by presenting known examples from evaluation of teachers 2
performance and the results of their ranking. Then we will examine the results by presenting new examples. A typical neural network with two hidden layers shown in figure 1. Figure 1: Artificial neural network with two hidden layers. Theoretical foundations and research history During the last year, many researches are done about performance appraisal and educational evaluation. In each of these researches necessity and merit of evaluation methods reviewed and attempted to improve the credibility and effectiveness of methods. These researches indicate that faculty members, students and researchers are interested in increased accuracy and efficiency of evaluation methods (Adem & Esra, 2007). Shaverdi et al (2013) presented an evaluation model based on fuzzy logic to evaluate efficiency of private banks of Iran. Adem and Esra(2007), presented a fuzzy model to evaluate and select sample employee by considering degree of competence and jurisdiction. They used analytical hierarchy process to attend influencing factors and criteria in evaluation and selection of samples. Yalcin et al(2012), presented a new approach to evaluate financial performance of productive industries in Turkey, by using fuzzy multi-criteria decision-making techniques. Hwang and Lai(1994), Shahrezaei(2010), Mirfakhradin, Owlia and Jamali (2009), Presented fuzzy decision-making models with multiple targets. The goal of these models was presenting performance evaluation methods and comparing the results. The main precise tool in analyzing and interpreting educational performance, is using mathematics and statistics. Yuzainee, Mohd and Azami(2012), described a fuzzy symmetric model as a replacement method for analyzing and interpreting performance of graduates from every skills of graduation. Method One of the main weakness of Danielson s (1996) importance of each member is not measurable in final decision. Otherwise, it is not easy to understand that which factor or member, affects managers decisions more than others? In the suggested method this problem is fixed and the importance and weight of each 3
member computed and ranked separately as variables in considered to achieve final goal. natural language. The following steps Classification of criteria based on Danielson s educational model Fuzzy reasoning system Coded criteria Artificial neural network Output (Rank of teacher) Fuzzy logic has many applications when it is combined with neural network computations. Indeed, fuzzy logic, neural network computations and genetic algorithm are the basics of the science of soft computing. In this study we applied Sugeno fuzzy inference system and membership functions and fuzzy if-then rules implemented by Matlab software version R2009. In the first step, we designed a questionnaire based on educational needs assessment in Danielson s (1996) educational framework for teaching model. Selection of teachers is done between sample teachers and by measuring their level of teaching. Considering teaching experiences and felt needs, teachers have been asked to prioritize criteria mentioned in teaching and evaluation based on Danielson s (1996) framework for teaching. In the next step, fuzzy set designed by considering priorities determined by teachers in questionnaire. According to results, from first 22 main criteria in questionnaire, 14 evaluated as main factors and these criteria determined by results of prioritization of teachers. Table 1, shows sorting out of scoring criteria. The scoring was based on teachers view about importance and prioritization of criteria in teaching different courses. Maximum score is 10 and minimum 1. Then, sum of all given scores by teachers calculated, then criteria with maximum score perched at the top of the table 1, other criteria in the next rows respectively. Table 1: prioritization of criteria according to Danielson s (1996) framework of teaching. 4
Row Danielson s Model Score (100) Fuzzification 1 Demonstrating knowledge of content and pedagogy. 92 Extremely important 2 Flexibility and responsiveness to work. 90 Extremely important 3 Establishing culture for learning. 82 Extremely important 4 Engaging students in learning. 75 Extremely important 5 Communicating clearly and coherently. 73 Extremely important 6 Using questioning and discussing techniques. 63 important 7 Designing students assessments. 51 important 8 Demonstrating knowledge of resources. 47 important 9 Growing and developing professionally. 44 important 10 Creating an environment of respect and rapport. 29 Mid-important 11 Reflecting on teaching. 26 Mid-important 12 Providing feedback to students. 22 Mid-important 13 Selecting instructional goals. 21 Mid-important 14 Managing classroom procedures. 20 Mid-important Fuzzification First 4 criteria which had maximum scores according to teachers view, considered as extremely important and classified in first fuzzy functions group. Other criteria, classified in three important, mid-important and low-important groups. According to teachers scores and ranking we removed 8 low-important cases because of their low effects on results. These fourteen criteria considered as values of first fuzzy set in this fuzzy reasoning system. The fuzzy set determines importance of each criteria in final measurement. 5
Figure 2: first input function of fuzzy set: importance of each criterion. To express criteria values, a fuzzy membership function designed and this function implements values of each criterion in fuzzy model. The reason of using fuzzy logic to create input and output membership functions is that the language of fuzzy reasoning is so close to natural language then it could be expressed and implemented easily by natural language measuring patterns. Indeed, too express measured criterion in a measuring sample, we used Extra good, very good, good, Medium and Weak phrases as fuzzy functions. Figure 3: Second input function of fuzzy system: fuzzy function of each criteria. The remarkable point is that the results of each criterion will affect final result. So, a fuzzy output we defined a Sugeno type of fuzzy output membership function to send every defuzzificated criterion as an input for artificial neural network after normalization. After answering to each criterion and by considering its importance. It will be measured through a set of fuzzy rules. It should be noted about fuzzy reasoning rules that these rules should be designed in a mode to be capable to consider all input forms and determine suitable output. 15 rules designed for fuzzy reasoning system, according to expertise of sample and experienced teachers. These results shown in Figure 4. 6
Figure 4: Fuzzy if-then rules Fuzzy reasoning system of this study reads two mentioned input and creates the output based on fuzzy reasoning rules. A degree of importance considered for each criterion and by that degree, it will affect final measurement. Figure 5: Fuzzy reasoning system. 7
Figure 6: example of firing of fuzzy rules and showing the output. Applications of Artificial Neural Networks for Classifying Teachers We used a perceptron multilayer network with a hidden layer. Number of inputs are equal with number of feature vector dimensions which it is the criteria values. It is necessary to convert nominal variables to numerical values to create a neural network model. On the other hand, it is necessary to normalize all input to (-1,+1) range. We normalize achieved outputs from fuzzy system before entering to neural network. We set different values for number of neurons in hidden layer and then calculated accuracy of each classification. There are 4 outputs for network and these outputs are classification of teachers in trainee, amateur, professional and advanced classes. During network evaluation, maximum output of network indicates label of class of input. In this study, we used backpropagation algorithm with Levenberg-Marquardt function to train the network. We set epochs of algorithm 1000 and mean square root 0.01. To test and train the network, we needed to evaluated samples. Required sample collected from interviewing and questionnaire from managers of 4 schools in 4 th educational and training area of Karaj city in Iran. Statistical population of this study include 106 of teachers of these schools. Teachers ranked using Danielson s (1996) framework for teaching after they were scored by managers. To train and test the network, 70 percent of data used to train network, 15 percent to test and 15 percent to validate the network. Results of running different neural network show that by increasing number of neurons, the accuracy of classification will be increased. If we calculate average of classification accuracy of different running, the accuracy will be increased. Finally after several training and testing of the model, we calculated average accuracy of classification 98 percent. 8
Figure 6 : A training neural network Conclusion To have qualified teachers in educational system of every country it is necessary to identify teachers strength and weaknesses, develop their professional skills, provide continued trainings in accordance with changes in information technology. A main approach to achieve mentioned objectives is continued evaluation of teachers in many countries. There are several models to analyze educational activities of teachers. In this study teachers ranked based ion Danielson s (1996) framework of teaching using fuzzy methods and artificial neural network. Indeed combination of soft computing paradigms causes to increasing accuracy and flexibility during projects. In the proposed method, first we converted measured criteria of Danielson s (1996) model to fuzzy verbal variables and natural language by presented classification of experts. Then it processed by fuzzy system and in each game membership degree of each criterion extracted by fuzzy rules and coding their outputs thus inputs of neural network will be provided. Neural network is the main part of learning pattern recognition system. For neural network, first we input training data to the system and the data are output of fuzzy system. After training the network, our created model will be evaluated and its accuracy will be measured. Results show that because of using software environment and artificial intelligence techniques, the created model has more accuracy, velocity and flexibility to ranking and performance evaluation. This model evaluates teachers performance based on interviews and questionnaire with education experts. It uses 14 main criteria of Danielson s (1996) model for performance evaluation. Finally, this model could be used as base for exact 9
educational activities, deserve to achieve educational degrees, professional development, ranking and paying salaries and benefits of teachers educational system. References: Adem, G., & Esra, K. (2007). A fuzzy model for competency- based employee evaluation and selection. Computer and Industrial Engineering, 52, 143-161. Avalosa, B., & Assael, J. (2007). Moving from Resistance to Agreement: The Case of the Chilean Teacher Performance Evaluation. Educational Research. 45(2006): 254-266. Danielson, C. (1996). Enhancing Professional Practice: A framework for Teaching. Alexandria, VA: Association for Supervision and Curriculum Development. Danielson, C. 2001. New Trends in Teacher [6]. Evaluation. In: Educational Leadership, 58(5), 12-15 Hwang, Ch. L., & Lai, Y. J. (1994). Fuzzy MODM. New York: berlin Heidelberg. Kimball, S. M., & Milanowski, A. (2009). Within a Standards-Based Evaluation System. Educational Administration Quarterly, 45:1. Mirfakhradin, S.H., Owlia, M.S., & Jamali, R. (2009). Reverse engineering quality management in center learning higher. Quarterly Journal of Research and Planning in Higher Education, 15, 131-157(in Persian). Mangiante, E. M. S. (2011). Teachers Matter : Measures of Teacher Effectiveness in low-income Minority schools. Educ Asse Eval Acc. 23, 41-63. Shahrezaei, S.R. (2010). Review evaluation in learning higher iron with presentation a performance. Quarterly Journal of Research and Planning in Higher Education, 16, 41-60 (in Persian). Strong, M., Gargani, J., & Hacifazlioglu, Ö. (2011). Do We Know a Successful Teacher When we See one? Experiments in the Identification of Effective Teachers. Journal of Teacher Education, 1-16. Tuytens, M., & Devos, G. (2011). Stimulating Professional Learning Through Teacher Evaluation : An Impossible Task for the School Leader? Teaching and Teacher Education, 27(5), 891-899 Shaverdi, M.R. Heshmati, E., Eskandaripour, A.K., & Akbari Tabar. (2013). Developing sustainable SCM evaluation model using fuzzy AHP in publishing industry. Procedia Computer Science. 17, 340 349. Yalcin, A., Bayrak, daroglu., & C. Kahraman. (2012). Application of fuzzy multi-criteria decision making methods for financial performance evaluation of Turkish manufacturing industries. Expert Systems with Applications, 39(1), 350-364 Yuzainee, Md Yusoff., Mohd, Zaidi Omar., & Azami, Zaharim.(2012). Evaluation of Graduates Performance using Fuzzy Approach. Procedia - Social and Behavioral Sciences 102, 64 73. 10