Machine learning is one of artificial intelligence subsets. The machine learning algorithm uses a calculation method to directly acquire information and learn from data without leaving a predetermined expression, and set "intelligent" to receive the machine as a model. Setting data received according to training data to construct an algorithm model based on the concept of machine learning is a specific design prediction by a progressive approach of I / O data applicable to the design. For large data, set a real-time approach. Powerful machine learning approach, vector machine (SVM) support (SVM) for data mining and pattern recognition problems, and one of ML software computing technologies is an artificial neural network Is known. Ann is an group interconnected with a neurons (artificial) using a computational model implemented through mathematical models. Three parts of the neuron network. The first part is the neural network connection pattern between neurons, architectures, or models. The second part is the procedure for determining connection weights, training or learning algorithms. These weights represent the information processed by the neural network. The third component is the activation function, which is the function that determines the output of each neuron. The activation function usually maps each real input to a finite range usually 0 to 1 or -1 to 1. This activation performance is sent to several other neurons according to the connection diagram seen in the architecture by only one of the neurons. This paper presented basic mathematical algorithms to connect the fuzzy logic in neural network , numerical calculation gives the result based on both in forward network and recurrent fuzzy group it will be most applicable with medical application. The goal is to take the simplest neural network applied for linearly separable patterns based on classification SVM algorithm. The data dependencies and data independence gives the effectiveness of a support vector machine learning algorithm approach of non parametric methods to tractable for massive datasets in feature of high dimensional.