The Study Of Service Recommendation Model Based On Mobile Users’ Web Access Content

The release of 4G licenses and sustainable growth of mobile terminal holdings inevitably lead to high growth situation of the mobile Internet. This also marks a new era of China mobile Internet entirely. For operators in China, the growth space of the saturated voice communication market is narrow. In order to promote strategic transformation, operators put forward to three main driving forces which are the stock management, traffic management and guest management. Through the information mining to customer, fine management, differentiated services, so as to better meet the needs of users and make the business continues to grow. How to get users’ potential spending power under the trend of the development of the mobile Internet is an important subject in front of operators.This thesis starts from the analysis of service recommendation based on users’ web access behavior and content in domestic and overseas. To use Internet access log of the mobile terminal for finding people’s behavior so that knowing the user’s life demand. The goal is realizing ‘product searching for user’. Then build the model of service recommendation based on web access content. Main contents and achievements of this thesis are as follows: 1. the preprocessing and analysis of user’s web logAt first, the thesis collects and filters a large number of users’ access log. Making statistics of online behavior such as page views(PV, page view) over a period of time to analyses time-preference of users’ Internet behavior. The second, web crawlers have been used to scrape the source information of pages. At last, this thesis uses text extraction technology to extract the text information in HTML. 2. the text classification of the user access contentUsing SVM(support vector machine) text categorization technology to get the class of users’ interests, then determine the users’ demands in this period, extract the target users. This thesis realizes the Chinese text participle, feature extraction of text, the texts’ vectorization and multi-classification based on SVM. Collect training sets to train classification model, then use the model to classify the users’ web access content. 3. product recommendationTaking maternal and infant product as an example, this thesis analyzes the features of maternal and infant products, then collects texts to construct feature vectors of different types of products in this filed. Extracting users whose preference coefficient in maternal and infant information is not zero as the target user, then calculate the similarity of users’ browse content and feature vectors to determine which kind of products will been recommended to user. At last, recommend products to people through time-preference of the people.