Learning and analysis on large scale data sets is very important in data mining. Large amount of data can be a cause of problem in learning time and also in learning capability. This research proposed a novel method to solve that problem by using multiple neural networks to learn from multiple sub datasets that extracted equally from the whole dataset. We employ algorithm by applying Euclidean Distance to integrate weight of hidden nodes from each network.The experimental results show that reducing the learning time for the proposed method and also preserve the accuracy.