Hidden information recognition based on multitask convolution neural network

With the continuous application of big data technology, machine learning is playing an increasingly important role in enterprise applications. User information, such as gender, age and educational level are the core factors for the research and application in the field of computer psychology, personalized search and social business promotion. This paper proposes a method for automatically inferring user information based on the search terms of users. We establish a multi task convolution neural network model based on word vectors. After the process of data cleaning, user search word segmentation, we use the word2vec to transform words into vector representation, and then build a multi task convolutional neural network model. This model is compared with other models based on word frequency, LDA methods. Experimental results show that our proposed multitasking based convolutional neural network approach works better than other methods.