結合深度學習及關鍵字搜尋熱度趨勢於臺灣鋼筋價格漲跌幅之預測

The fluctuation of the rebar price is hard to predict due to too many factors involved within the production of the steel industry. Current models and database for predicting the rebar price mainly focus on the economic performance, which is not suitable for construction industry. Along with the rapid application of the artificial intelligence technology, the deep learning concept is widely used in developing algorithms for predicting model. Convolutional neural network, the class of the deep learning, is good at extracting features from multidimensional data. Therefore, this research develops a convolutional neural network to predict the fluctuation of the rebar price. First, the factors including domestic and foreign events that could impact the rebar price are analyzed. Then the data of the above-mentioned factors are transformed into picture to train the proposed prediction model for rebar price. Finally, keywords that represent different events impact rebar price are used to calibrate the accuracy of the proposed model.