Microblog Big-data Analysis Based On Multimodal Learning With Deep Boltzmann Machine

Deep learning is currently an extremely active research area in machine learning and pattern recognition. It has gained great successes in area of Speech recognition, computer vision, and nature language processing. Under the big data society, it can become to the big data analytics driver. Bigger makes more accuracy. The most important part of deep learning is Boltzmann Machine, Which include configuration of Boltzmann Machine as a single layer feedback network and classification of Boltzmann machine according to the topological structure, including General Boltzmann machine, Semi-restricted Boltzmann machine and restricted Boltzmann machine.Analysis of the human mental state from microblog data to perceive human mental stress has attracted widespread significant attention in recent years. This state is in contrast to many existing methods that mainly focus on the textual modality data and ignore the effect of image modality data that also expresses the human mental state. Through multimodal microblog data to analyze human mental stress, the challenge is the very difficult learning of cross-modal features from low-level representations. To solve this problem, the paper presented using multimodal learning algorithm with Deep Boltzmann Machine (DBM) to analyze microblog image and textual data, the modal can implement to transform low-level features of images and texts to sparse high-level abstract concepts, and then a joint representation layer is employed to fuse common features derived from the two different input modalities. In addition, the model can cross the two different modalities to discover the highly non-linear relations in low-level features. The experiment results show that the proposed method is effective.