Mining mobile information service patterns based on multi-view features fusion

Various mobile applications (Apps) have overlap functional features, a large number of user reviews, and multiple labels, which may cause a difficulty for market opportunity discovery, application integration and selection. This paper proposes a text mining and network fusion analysis framework for finding information service patterns. First, similarity networks are built from three views of function description, user reviews and labels. Then, different similarity networks are nonlinearly fused in an integral manner. Data of 2451 mobile travel applications were crawled, three networks got fusion to form a comprehensive view, overcoming different measurement and noise, taking advantage of complementary information. Fused networks were used for clustering. External evaluation result had significantly been improved including normalized mutual information and accuracy. Mainstream travel service patterns were found, such as map navigation, train and car tickets, taxi car, bus and so on.