| Nowadays in the mobile Internet era, people are often submerged in the raging torrents of information. With the rapid development of Network news and information, people's work and life paces are faster and faster. In order to make better use of fragmented time to complete the acquisition of news information, the news media people read needs to have more accurate, faster and more personalized differentiated news information focusing ability. Therefore, personalized news recommendation systems increasingly mature. All personalized news recommendation systems are able to provide information in line with the original intention of their interest preferences for different news users. Nowadays in a variety of news recommendation systems based on different strategies, the thesis mainly studies news topics classification, cold start, data sparse, user interest model and other issues in the news recommendation systems, and proposes a personalized news recommendation method based on news text automatic classification and mixed recommendation, which mainly includes the following three points:1. News text content topic classification is analyzed, and news text classification and word frequency content mapping method are used to establish the feature category sequence of news text news, which is the news text feature representation.2. A hybrid of content-based and collaborative filtering personalized news recommendation method is proposed. This method improves the traditional content-based news recommendation method. The user interest model based on the content-based method and the user interest model based on collaborative filtering method are mixed in the method, and the background factor of time context is also added, and then a new user interest model is obtained. By calculating the similarity of this new hybrid interest model and news text vector,it's able to recommend personalized news which make users interesting for different news users. At the same time, the method researches the novelty and diversity of recommendation news, which would effectively improve the novelty and diversity of news.3. Complete the design and implement of news text automatic classification system and the personalized news recommendation system based on the hybrid recommendation method. |
PDF Full Text Request