The Design And Implementation Of Product’s Ranking Of Merchandise Search Engine

With the development of Internet and e-commerce, the c2c and b2c’s sites provide users with more and more goods.But the online shopping sites only provide search service to users in their stations.A whole network search engine of merchandise will be needed when the user want to compare the goods which is searched is on sites. Merchandise search engine of the whole network is a collection of all businesses, commodity information, and it is a vertical search engine which can provide a variety of dimensions to retrieve.Merchandise Search engine provide the function of retrieving merchandise in the whole network, if the information returned by the query in the Merchandise dimension is presented to users, it will allow users submerged in a large amount of data. A variety of methods in improvenment of the display of the results of inquiries has been proposed.Which returned results up from the commodity dimension reduction (Reduce) to the product dimension to the show are a better approach.Product is a generalization of the concept of the commodity.The search of users can be navigated to the dimension of product and then compare and select goods which belong to the product. Therefore, the authority of the product’s ranking in a certain extent reflects the authority of the sort of Merchandise search engines.Based on the above background and combined in the work of the internship company, I designed and implemented a module of product’s ranking in a Merchandise search engine.The static socre which is produced by the model of product’s ranking as the basis of rangking in Merchandise search engines has been worked online.In this thesis, product ranking module is divided into two sub-modules:Product static scores calculated off-line module and data monitoring module. Static product score calculated off-line using Hadoop technology, so that it can handle vast amounts of information and data. The module is designed with scalability, several functions is provided, for instance it makes changes to the calculation of standard products in different categories according to operation assitant’s requirements and can process the product node which is unusual. Data monitoring module which can to monitor the function of the product fraction fluctuations and product’s features, is provided to the developers. The developers can read the output reports to intuitive observation or chase the reasons of abnormal scores. The module uses the Django framework, combined with Django’s MTV development pattern; it is divided into template layer, the view layer and model layer from top to down. The separation of the template layer which is presentation layer and logic processing layer which is view layer, the model layer, allowing developers to more easily develop data-driven web application.In this thesis, first of all, the background of the project is introduced. And then, the technologies and frameworks which are used in the project are brief introduced. After that, the demand of the project is analysised. On the base of the demand analysis, the design and implementation of the sub-moduels of the static score’s producing and data moitoring are elaborated. In this section, the mapreduce program of the calculating of the product’s static score is the most important. Finally, the summary and outlook of the project are made.