An improvement of artificial neural network and the comparison with the previous

摘要

Recently, artificial intelligence plays a more and more important role in our study and daily lives. People use it to forecast or make the best decisions. Artificial neural network (ANN) is the most important model in the intelligence forecast. However, the model is not satisfying enough that the accuracy is just 80%–92%. So we need to strengthen the model to make it do a better job. In this paper, we come up with a way to improve the artificial neural network, through which users can do forecasting, classifying and other work more exactly. To implement this model, we add a new parameter to the activation function so that the whole model may function more accurately. And at last, to test the improvement of the model we have got, we do an additional experiment in order that the model makes the accuracy rise by 0.5%.