Text Classification Algorithm Based On RBM

With the advent of the era of big data, the explosive growth of information isin the form of an annual index. Select the information we want from these data seemto become more and more difficult. Fast and efficient selection of information fromthe mass of information we want, and to classify these vast amounts of informationand management has become a matter of growing concern. Data mining is animportant area of machine learning, it can in a certain extent to solve the aboveproblems that plague us. The text classification is an important application of datamining, which can quickly help users select the information you want, and be able tohave the information classification and management, has great significance.RBM (Restricted Boltzmann Machine, RBM) is a neural network model basedon statistical mechanics. In recent years, with the rise of the field of deep study, as itsbasis the RBM model has also been widespread concern. But most of RBM regardedas feature extraction methods or basic parameters for the neural network to obtainrationalization, few of its classification as an independent study, in this paper, wedescribe the process of RBM as an independent classifier classification and proved byexperiments, it performed well in the classification performance. Which can be wellhidden in the learning text information, its utility value is also high.In this paper, the choice of two different feature selection algorithms andcombining RBM classification; emphasis describes two feature selection algorithmsand RBM classification process by RBM classification experiments to verify thecorrectness and efficiency; RBM is a very good neural network model It can simulatethe information we want, is worth a serious effort to research and mining.In this article, we focus on content and innovation are as follows:1. The studyof some theoretical text classification to be used throughout the text of the generalclassification process; including text preprocessing, feature extraction andclassification algorithms to choose the right training, then preconditioning test text, and finally the use of training good classifier for classification and prediction.2. Thenfor some classic classification algorithm is studied, summed up the advantages anddisadvantages of their existence.3. Classic RBM model was carefully studied in depth,including the network structure RBM model, the energy function and probabilitydistributions and related RBM training algorithms.4. A novel feature of wordselection method, it is a class exclusive word feature selection algorithm, using itsclassification algorithms RBM process were studied.5. Made a word-based ExclusiveRBM classifier and frequency of RBM-based document classification, by experiment,a comprehensive analysis of their performance; then categorized according todifferent evaluation criteria, made the corresponding comparative analysis.