Simulation of Big Data Order-Preserving Matching and Retrieval Model Based on Deep Learning

Although the speed of big data processing can be effectively improved through advanced infrastructure such as cloud computing and supercomputing and parallel distributed processing means, it is extremely important to seek an accurate and fast matching and retrieval algorithm for improving the order-preserving matching and retrieval accuracy of big data applications. This article is based on the DL (Deep Learning) big data order preserving matching and retrieval model. By implementing a layer by layer initialization strategy for encoding, the stack style self coding neural network is initialized, and the activation state of the hidden layer neural unit is used as input for the next level. The DBN (deep belief network) with RBM (Restricted Boltzmann Machine) as the basic unit is trained. Based on the distribution of text themes and image feature vectors, big data ordering matching and retrieval are achieved. The simulation results show that for the same random data series, with the increase of pattern length, the number of matching results will gradually decrease, and the proportion of similarity calculation time to the overall time of the algorithm will gradually decrease. In multi-data retrieval, the proposed method shows the trend of more data and higher accuracy, which has obvious advantages.