ANN is a machine learning method. It can be used for pattern recognition and prediction of multi-variant time series. The training performance of an ANN depends on its architecture and the training data. ANN architecture is inspired by biological neural network of the human brain. The classical feedforward ANN architecture includes an input layer, a number of hidden layers and an output layer. Each hidden layer includes a number of parallel distributed hidden neurons. Compared to other machine learning methods, the advantage of ANN design is that an ANN can be trained without knowing the features of the training data beforehand. The disadvantage is that it requires a large number of training data to obtain good ANN training performance. Forecasting using ANN can be dated back to two decades ago (Zhang, 1998) and has been successfully adopted for time series pattern recognition and prediction in many applications. Recent publications have shown promising research advances in optimizing ANN architecture and training algorithm for time series prediction (Zhang, 2018).