Research And Implementation Of An Integrated Simulation Platform Recommender System Assessment Framework

The rapid development of Internet technology has led to the problem of information overload. Faced with the problem of information overload, although related applications, such as retrieval systems and search engines can help people find more accurate information. In some applications, such as movies, music, shopping and dating platform, users generally can not make a good retrieval good demand, and in some cases, put forward by the user to obtain a search keyword for users terms may be worthless. Recommender systems solve the above problems, it is through the history of research user, the user’s social information and attribute information of the corresponding classification of data, information or classified data modeling resource modeling user through some reliable way the way the user of potential interest to recommend to the user data resources. For just adding users, some websites using questionnaires and other means to add more details to understand the user’s preferences and needs, Initially, the general use of social recommendation and content-based recommendation algorithm. After receiving feedback from users of long-term data, the formation of a valid user data sample is recommended to start using other algorithms, such as collaborative filtering algorithms. In addition, the content provided sites of some specific areas require users to join the site, Before using the service provided by the site, the user must provide detailed information on user tags and scoring perfect for some commonly used classification of information and data, and thus provide a more accurate recommendation results.This paper presents a design to achieve a comprehensive framework for assessing a recommendation system simulation platform that can simulate a variety of input data sets to provide input to the algorithm, and these data sets unified management and storage, and its data processing, for different algorithms are hot-swappable, the corresponding algorithm can be structured to comply with different configurations to use the system before the unified management of data sets, thereby offline test, then the results of the assessment results obtained according to the feedback, so that the algorithm provider for algorithm can have a better improvement. This article looks at the recommendation system simulation platform integrated assessment framework datasets unified management and collection, diverse data sampling, data cross-validation, data visualization, data characteristics and recommended assessment results show, for the concrete implementation of the algorithm and the recommended results return, is not the focus of this article, this article will be a simple example shows the user can freely configure the algorithm execution thereof.Idea of this paper: first recommendation system development status are summarized study to build a recommendation system requires the use of an integrated simulation platform to the related technologies, discusses common data processing techniques, clustering algorithms, data visualization, and distributed storage technology. Then discuss the specific needs assessment framework, followed were in accordance with these requirements in data entities container frame design, data presentation tools and features for simulation and evaluation module, choose to use django MVC frameworks such as the user interface, and finally achieve such diverse comprehensive simulation and evaluation platform for data entry.