| With the rapid development of electronic commerce, Internet advertising and merchandise is more and more, at the same time, the user gradually is to the ads produced like " immune", resulting the CTR sharping decline in recent years. So for display advertising, network operators, advertising agents and advertisers hope to improve ad click rate, thus earning more advertising revenue or the formation of more potential sales revenue, thus to find out the influence of CTR key-factors and take targeted measures is so important.Based on the Chinese word segmentation, topic extraction and gray to forecast a model core algorithm, on the online shopping website user behavior analysis is studied. In the paper, based on the specific user behavior log analysis processing, the user may be interested in topics related to commercial or advertising, which targeted to specific users by correlation degree higher advertising push or for particular advertisements to range higher user push, its core target is improving CTR forecasting results, to make enterprises get more profits.The basic flow of the design is based on the user’s accessing histories, put forward a kind of user behavior model, and use the grey prediction model for prediction, the user will click results; at the same time, through a certain topic extraction based on user history data and advertisement data subject extraction; and then using the vector space model to forecast results with the advertising theme correlation computing power, last value( correlation) of the order from high to low are arranged, in order to complete the ad / commodity push.The design of the final results will be added to the bigger e-commerce systems or advertising push system, in a modular form together to complete a series of functions. |
