pure, shared
Table of contents
2 Dynamic evaluation model of data price
2.1 Data Product Pricing Strategy
Summary
In traditional data trading platforms, pricing is completely controlled by the platform, data owners do not know the potential value of data, and online buyers and sellers lack credit, making it difficult to evaluate data prices in data transactions. This paper proposes a dynamic evaluation model of data price based on credit game. The model includes: data product pricing strategy and credit game model. Buyers and sellers engage in a credit game through the price asking and quotation mechanism, and the data price is obtained through the credit function obtained through the credit game model and the data product pricing strategy. This paper proves the existence of mixed-strategy Bayesian equilibrium and mixed-strategy perfect Bayesian equilibrium in basic game process and repeated game process of credit game model respectively. Finally, through simulation experiments to simulate the transaction between buyers and sellers, it is concluded that the credit function plays an important role in data transactions. This model can complete the data price evaluation, the price difference percentage is controlled at about 10%, and the transaction success rate reaches more than 96%.
In real life, commodities are valuable, but when the production cost of commodities is difficult to measure, it is difficult to evaluate them. People use deep learning methods to evaluate them through feature learning , such as the application of neural networks in the evaluation of used car prices [1] , A financial market trend prediction method based on sentiment analysis [2] . However, these price evaluations only perform feature learning on a single product-related data set to complete the evaluation.
Data is different from real objects. The forms of data and the content of data are various, which makes it difficult to learn features. Therefore, we should give full play to the functions of the market, return to the original price formation mechanism, and give full play to the initiative of buyers and sellers.
The main contributions of this paper include two aspects:
- Establish a data price dynamic evaluation model , including data product pricing strategy and credit game model [3 ] . The dynamic game process is analyzed and the existence of Nash equilibrium is proved . The credit game model solves the problem of dishonesty between buyers and sellers; the pricing strategy of data products prevents the deviation of data value;
- 11
1 Introduction
The era of big data has arrived . According to the "2021 White Paper on the Development of China's Big Data Industry", the scale of China's big data industry will reach 638.8 billion yuan in 2020, a year-on-year increase of 18.6%. By 2025, the scale of China's big data industry will reach 3 trillion yuan [4] .
With the gradual maturity of machine learning, deep learning, neural network and data mining technology. The role data plays in these technologies is increasingly evident. The development of data transactions can bring benefits to data providers on the one hand, and on the other hand, data buyers can also meet their actual needs. To achieve mutual benefit and win-win results for both parties.
The price of data and its cost are seriously out of balance . Data has a fast growth rate and its value is difficult to estimate [5-6] . In the traditional central server data trading platform, users will lose control over the data when they upload the data, and the data pricing is simply set by the seller ( object , O ) or by the platform. The value assessment of data and the authorization of data transactions rely entirely on a third-party central platform , resulting in a serious mismatch between data prices and costs . There is no bargaining involving the dynamics between buyer ( subject , S) and seller O. Judicial positioning and protection of data [7] proposes to set absolute rights to data files, and the original acquisition of data file ownership. From the perspective of artificial transaction concepts , data file producers use legal weapons to protect their rights when their data rights are violated by others. . However, this method is no longer applicable when the value of data is lower than the cost of litigation.
Trust relationship is one of the complex social relationships of human beings, and a good trust model can resist attacks [8-9] to achieve credit evaluation . Literature [8] defines a two-way trusted transaction protocol between two different assets , divides the transaction process into three states to identify the transaction stage , and realizes the exchange of different assets without increasing computer resources . This article uses the game method to let buyers and sellers choose to ask or bid for a credit game . Therefore, the dynamic evaluation of data prices in an open trading platform is an incomplete information dynamic game [10-11] between the buyer S and the seller O. In mobile cloud computing, based on dynamic game [12], the recommended incentive strategy of dynamic game can improve the data security and privacy protection of mobile cloud services. We apply this incentive strategy to credit evaluation.
In an open trading platform, after the initial request behavior of buyer S to seller O , the request behavior may occur again after a certain time interval, that is, the former is a basic game, and the latter is a repeated game. Repeated game [13-14] means that the game with the same structure is repeated many times, and each game is called "stage game". Repeated game is a kind of multi-stage game. It is a simple repetition of the basic game. The strategy variables and payment structures of the two players in each stage are exactly the same. Therefore, the repeated game process is considered to be composed of basic static games [15 ] . . However, in the process of interaction between buyers and sellers, the actions of both game entities after each "stage game" are observed by both parties, and the observed results will directly lead to the selection of behavioral strategies and equilibrium results when the game behavior occurs again. Therefore, Repeated game is also a dynamic process, which belongs to the category of dynamic game, and repeated game cannot be simply regarded as a linear superposition of basic games. Obviously, the repeated game between buyers and sellers has the dual characteristics of open structure and closed structure: (1) the information structure of each stage of the repeated game is not affected by the previous behavior of the other party, so the strategy of the player has the nature of an open strategy; (2) Since both the buyer and the seller have observed the past historical behavior of the game, the buyer S ( seller O ) can take corresponding actions against the behavior of the seller O ( buyer S ) to guide the future actions of the buyer S ( seller O ).
Because the utility of repeated games is different from single static and dynamic games, it is not a total utility after the entire repeated game is over, but the utility generated in each "stage game" including the game process. Therefore, for an open trading platform, this paper builds a dynamic evaluation model of data prices in data transactions based on trust , and gives the extended and payment matrix of the trust-based request-control game [16-18] . Through the analysis of the interaction process between the two sides of the network entity, the existence of the basic game process of the trust-based dynamic request control game and the mixed strategy Bayesian equilibrium and the mixed strategy perfect Bayesian equilibrium [19-21] in the repeated game process are analyzed respectively . proof.
2 Dynamic evaluation model of data price
This model consists of two parts: data product pricing strategy and credit game model. The data product pricing strategy prevents data value deviation ; the credit game model obtains the credit function value of both buyers and sellers, and dynamically evaluates the credit of buyers and sellers . The credit function value obtained through the credit game model is combined with the data product pricing strategy to complete the dynamic evaluation of data prices .
2.1 Data Product Pricing Strategy
Data product pricing strategy: The platform side will obtain the selling price according to the nature of the data, such as the degree of scarcity, the size of the data volume, and the number of data items , and calculate the recommended selling price SP . Then obtain the price range [p1, p2] according to the price range table.
Among them, the degree of scarcity is divided into three levels, namely low, medium and high. The degree of scarcity is divided by searching and comparing the keywords submitted by the seller with the previous data transaction records. Determine the appropriate rank based on the number of entries retrieved. If the number of entries searched by keywords is within the range of three to ten, the degree of scarcity of the data set is considered to be medium; when it is less than three, the degree of scarcity is considered to be high; if it is greater than ten, the degree of scarcity is considered to be low. Scarcity is Medium Set the value of the scarcity level to one . The specific classification is shown in the classification table.
