Studies on the consistency of fuzzy regression models

In today's World, collecting data is not a big issue in a proper and processable format with the help of recent technological developments but it is a big deal not only for the government bodies, private sector and also for the individuals. The main idea behind keeping data in a processable format is to discover the existing data patterns to estimate the future with the highest consistency. Enhancements in data collection, processing and estimation technics bring the strong knowledge from the past, ability to control processes and dominate the future based on the historical data. In other words, technical control on data management promises the power of knowledge to the all stakeholders in management and in many areas mainly in economics, engineering, and medical sciences as well as politics and public relations. One of the leading research area in data management is modeling which are namely stochastic models, statistical models, lineer modeling or fuzzy systems. Fuzzy Models are one of the most trendy and consistently estimating approach for modeling. In this study, the data set of Denmark which contains financial indicators as input and import amounts between 1993 and 2013 as output were used to create 221 different models by using Fuzzy C-Means Clustering Algorithm and Fuzzy Regression Analiysis with different number of clusters and degree of fuzziness in each model. With the help of high number of experimental models, evaluation of model consistency parameter, R-square, depending on the changes of number of clusters and degree of fuzziness have done and several inferences have been achieved. Last but not least, outstanding approach has been developed in this study which is named as Multi-Layer Fuzzy. In this approach more than one membership matrix calculated with at least two different number of cluster and degree of fuzziness values have been used in a single model and higher consistency in the estimations have been achieved as expected