Shulou(Shulou.com)06/03 Report--

Information quality model is also widely used in the Internet industry and Internet data operation. Specifically, the most direct and critical link between the e-commerce industry and the e-commerce platform between buyers and sellers is the massive commodity catalogue, commodity Offer, commodity display, etc., whether it is B2C (such as Dangdang, VANCL), C2C (such as Taobao), or B2B (such as Alibaba), as long as it is for commercial purposes and for trading purposes. All need to adopt effective means to improve the quality and structure of massive commercial information (catalogue, Offer, commodity display, etc.), so as to promote transactions.

Under the same conditions, an online store or commodity display with complete elements, reasonable layout and friendly interface must be easier to reach a transaction and gain the favor of buyers than those without core elements, unreasonable layout and unfriendly interface. What is revealed here is actually the important value of information quality.

The application of the information quality model in the Internet industry mainly includes commodity Offer quality optimization, online store quality optimization, online forum posting quality optimization, filtering optimization of prohibited information, and so on. All scenarios involving information quality monitoring and optimization are applicable to (or draw lessons from) the solution of the information quality model.

The construction of information quality model mainly involves conventional data mining techniques, such as regression algorithm, decision tree and so on. However, for the requirements of the information quality model, because its target variables have certain particularity, there will be some differences in ideas and methods between it and the target customer prediction (response) model, the specific contents are as follows.

Any model is built to respond to specific business scenarios and business requirements. Sometimes the target variable of building an information quality model is whether the information (such as commodity Offer) generates a transaction in a specific period of time. In this case, the target variable is binary, that is, yes or no.

More often, the target variable of the information quality model is not directly related to the transaction (which is easy to understand because there are too many factors that affect the transaction), and sometimes the information quality itself is a subjective judgment, in this case, there is no clear target variable from the actual data. So how do you define the target variable?

According to the score of experts, model fitting is a more appropriate flexible strategy. For the specific operation of expert scoring and model fitting, the following takes the project of "star division of commodity Offer" as an example to explain and demonstrate.

Merchandise Offer is actually an online transaction in which the seller displays specific commodity details, transaction terms, picture details, etc., for each commodity sold, so that it constitutes a complete page. Generally speaking, after the buyer browses a specific commodity Offer, as long as they click "join the shopping cart", they can carry out the follow-up purchase payment process.

In a project of "star division of commodity Offer", the target variable was scored by experts, and business experts and industry experts scored artificially according to the weight of commodity Offer elements based on the professional background knowledge of the industry. These elements include the length of the title, the number of pictures, the proportion of optional attributes, whether there is a hierarchical price range, whether to fill in the total supply information, whether there is a mixed batch description, whether there is an operation description, whether to support online third-party payment and so on.

First of all, take a certain number of samples, ask industry experts to score and assign values to these samples one by one, after obtaining the specific scores of each commodity Offer, take these scores as target variables, use various models of data mining to fit the relationship between these elements and the total score, and finally form an appropriate model, which effectively integrates the opinions of experts and effectively fits the relationship between the constituent elements of Offer and the total score.

In order to be more accurate, on the basis of expert scoring, it can also be supplemented by customer research, so as to modify the expert scoring and the weight of various elements, and finally build and fit the model on the basis of revision, which belongs to the technical details of the project, not the core of the project, so we do not do in-depth explanation. The information quality model is the basic guarantee of e-commerce and online transactions, and its main purpose is to ensure the quality and efficiency of the basic information of goods, so that buyers can understand the main details of goods more easily, clearly and efficiently, and let sellers display their goods more easily and efficiently. Whether it is C2C (such as Taobao), B2B (such as Alibaba), or B2C (such as Dangdang, VANCL), we can use similar methods to optimize and improve the quality and effect of our product display, and effectively improve and ensure the conversion rate of transactions.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.