The Design And Implenentation Of Data Analysis And Visualization System For Accurate Advertising

Advertising is an important kind of profit modes for Internet service providers. It is a great challenge to catch user demands and deliver right ads to users translate so as to attract advertisers. From the perspective of consumers, most of advertise are annoying and even intolerable. In order to solve this problem, this article is in view of the search engine user data and its video website video service, with the analysis of users’ search records, user’s preferences, potential demand of the deep mining and real-time user perception. The implementation of precision advertising is aimed at mining the organic connection between consumer demand and advertising goods to build harmonious communication channels. Precision of target users of advertising not only can bring more to advertisers and advertising business interests, for users it reduces the harassment which brings a more convenient and quick guide for users’shopping and searching.There are mainly four challenges in this paper:the enormous crowd of fine-grained classification based on user behavior, the crowd model based on user feedback update, real-time update of dynamic data processing, a visual data analysis for ordinary users. Portrait based on user behavior modeling is the key to solve the first problem, according to the characteristics of the search behavior and video service, established a user-oriented search and play behavior, time and attributes of three major categories of nearly ten thousand dimension user characteristic vector and the organic integration of the user’s basic information and internal demand of external rendering. Through the study of the clustering of users and with the aid of the statistical results of each type of user characteristics, the system bases on the connotation of the user of the concept of mining to analysis of group characteristics and requirements deeply. In the light of the problem of model updating, system uses dynamic tracking and actions of the user behavior prediction by means of logistic regression for accurate advertising. For real-time data processing problems, with the aid of Hadoop technologies for huge amounts of data preprocessing and behavior prediction and using part of tuning in other business good non memory algorithm sparse user input and feedback of the tens of thousands of dimension data clustering analysis effectively classification processing, precision advertising, system can achieve advertising incremental updating in 15 minutes. In order to meet the demand of ordinary advertisers’user data analysis, system has made effective for massive latitude data division and integration and built established the industry overview, based on keywords and regional characteristic industry details and geared to the needs of the user’s interest and the information, gender data visualization platform.This paper describes the realization of the accurate advertising and data visualization system. First of all, this paper analyses the feature and shortage of traditional advertising, and Google, youku, CSDN in related industry advertising characteristics of Internet marketing service providers which makes it clearly about the precise advertising based on search engine data system of the superiority and necessity. Then, from the user information management, system data collection, keywords retrieval functions, audience segments and visual function requirement analysis of system are described in detail.In terms of architecture, front end uses play framework which is a lightweight web framework design. And the backend on distributed computing platform uses the python which uses the running hive SQL for user data statistics and architectural design of the algorithm based on graphs. Finally, we make detailed design and implementation of the system. The designed and implemented are based on K-means clustering algorithm and user click advertising prediction algorithm are based on logistic regression. And user data of user interests and gender orientation were described in detail. Through advertising feedback comparison before and after the system deployment we verify the effectiveness of the system. Compared with the same industry and the traditional way of advertising, the arrival of the click rate, one thousand advertising cost and average click prices have greatly improved. Run up to now, this system improves the user clicks in one day and brings almost doubled advertising revenue which is highly appreciated by the advertisers. The system also has important reference meanings.