Developing an artificial machine learning method based on behavioral learning

In the context of artificial machine learning many methods have been developed and actively used in analysis and evaluation of data. Some of these methods are the behavioral learning methods take human learning process as the bases for the method. Despite of that the results obtained with the behavioral learning methods depend on the experimental data, there are contradictory arguments between methods and they do not hold the same conditions, which can be understood from the literature. From the information available in the literature it is understood that more general methods in this concept are need to be developed. For this aim, among the popular behavioral learning theories like Pavlov, Watson, Guthrie, Thorndike and Skinner's conditioning theories were analyzed and animal based experimental studies were conducted toward the contradictory issues to remove these contradictions. In the light of the obtained experimental results, such parameters like continuously learning and forgetting, which hold in human learning, were taken into account and a more general behavioral learning method was proposed and the mathematical model for this method was developed. In this study it was demonstrated and shown that the developed generalized mathematical method is keeping on all of the properties of the available theories. Since the proposed method has dynamically learning, forgetting and back propagating talents, as was mentioned in the body of thesis, particularly it was proved that it is much successful in evaluating dynamical data compared to methods like ANN and kNN mentioned in the thesis. Furthermore, in order to show the usability of the developed method in solution of classification and regression problems, the model was tested on the update datasets. The obtained results showed that the developed model can be used in psychology (behavior control, phobia attenuation etc…), control, robotic and similar fields.