Product-Oriented Customer Targeting Mechanism

With the continuous development of the Internet, finding potential customers for a product has become a critical issue in many areas, especially as the arrival of Web 2.0. This paper aims to study the problem of product-oriented customer targeting mechanism. Different from the traditional customers’ discovery mode, our work mainly focuses on pushing the product to the potential customers:In our work, we propose a product-customer matching framework to handle this issue. In general, to complete the match between the user and the product consists of two modules:the first one is the generation of user interest and the second one is the product-driven query, which is also called the reverse ranking query. Hence, our framework mainly consists of two phases:data preprocess and query process. During the data preprocess, we devised a novel rule to generate the user interest. In the query process, we first design two novel queries, namely reverse k-Ranks query and reverse Top-k-Ranks query to find potential customers for a given product, which is critical in various applications, such as job-hunting and dating. The main contribution of the paper is as follows:·The rule-based user interest learning method. With the development of Web 2.0 applications, more and more users are attracted to generate reviews and ratings for the products which have been consumed. Our research scenario mainly includes life style platform, like dianping.com, Yelp and jiayuan.com, etc. Due to the data sparsity and short contexts in this platform, it is very difficult to build user interest preference based on the reviews. Fortunately, we can use the exact rating information to learn the user preference. This paper mainly employs the user’s ratings for the product to build user interest. We also extend the original linear model to describe the degree that how a user likes a product.· Reverse k-Ranks Query. In this paper, we first propose the reverse k-ranks query. Based the original linear model, this query returns k matching customers for a given product. We devise three pruning-based methods to answer reverse k-Ranks query efficiently. Tree-based pruning approach (TPA) eliminates some unnecessary pair-wise computations from the perspective of products. Batch pruning approach (BPA) reduces unnecessary pairwise computations from both perspectives of products and users. Marked pruning approach (MPA) reuses previous computation results of some buckets to further reduce the time consumption. We also devise two sorting-based methods in the two dimensional scenario to answer reverse k-Ranks query efficiently. Sorting-based algorithm (SA) improves the query performance by sorting the user preference. Except the sorting, Tree-base algorithm also builds a R-tree to further reduce the complexity. Thereafter, we conduct extensive experiments on real and synthetic data sets to verify the effectiveness and efficiency of the proposed methods.· Reverse Top-k-Ranks Query, we first propose a new query named Reverse Top-k-Ranks Query to find some users to match the query product. This query is based on the extended linear model and integrates the results of reverse top-k query and reverse k-ranks query, improving the coverage ratio of returned users. Then we devise two efficient methods to handle this new query, including the extended RTA (ERTA) and History-Based Batch Pruning Approach (HBPA). ERTA is based on the RTA algo-rithm and employ a threshold to reduce the computation complexity. HBPA employs a pruning strategy from the perspective of both products and users. The experimental results show the efficiency and effectiveness of our framework. Thereafter, we conduct extensive experiments on real and synthetic data sets to verify the effectiveness and efficiency of the proposed methods.In summary, this paper mainly study three critical issues for the product-oriented customer targeting mechanism, including how to build the user preference interest, reverse k-Ranks query based on the general linear model and reverse Top-k-Ranks query based on the extended linear model. Research on these three issues has continuity and sustainabili-ty, which offers a complete framework for the matching between customers and products. Our work is based on the comprehensive survey and analysis on existing theories and tech-niques. The theory analysis and extensive experiments show that the proposed methods for above four problems achieve good effectiveness.