According to the traditional recommendation technology which is applied in large-scale mobile data set is difficult to balance the accuracy and efficiency of recommendation, this paper puts forward a hybrid recommendation algorithm based on collaborative filtering and Word2Vec. The traditional collaborative filtering algorithm is realized with MapReduce framework and Hive database on the Hadoop platform. Use the Word2Vec model to train the tag information of the mobile data to get the similarity between the tags, and recommend applications to the user according to the similarity. Finally, according to the feedback behavior of the user, the recommendation results are mixed with weight. The experimental results show that the hybrid recommendation algorithm improves the efficiency and accuracy of recommendation greatly, and makes it more advantageous in the large-scale data set.