CN114445099A - Returned freight insurance pricing method, pricing system and construction method thereof - Google Patents

Abstract

The invention belongs to the field of return freight insurance pricing, and in particular relates to a return freight insurance pricing method, a pricing system and a construction method thereof. The return probability of the current product; the return probability is brought into the profit maximization objective function to calculate the return freight insurance price. Secondly, by using the optimal price obtained by the above pricing method as the price soft target, using the above behavior data and purchase data, the return probability prediction model takes the return freight insurance price as the prediction target, and calculates the model. The error between the predicted price and the soft price target , optimize the model parameters, and obtain the pricing system of each customer, which is used for the corresponding customer's return shipping insurance pricing. This method makes full use of the sufficient and large amount of customer behavior data and historical data owned by e-commerce companies, and can obtain optimal pricing decisions in one step without manual participation, and has high computational efficiency.

Description

A kind of return shipping insurance pricing method and pricing system and construction method thereof

technical field

The invention belongs to the field of return freight insurance pricing, and more particularly, relates to a return freight insurance pricing method, a pricing system and a construction method thereof.

Background technique

With the dailyization of online shopping, the online orders of e-commerce companies are also increasing, and according to statistical surveys, an average of three orders will generate a return order. There are many reasons why customers choose to return products, such as product quality problems, regrets after consumers' impulse purchases, products that do not meet consumers' expectations, and so on. Return freight insurance is a new type of insurance proposed by insurance companies and e-commerce companies to solve the losses caused by returns. Consumers can voluntarily purchase the return shipping insurance when purchasing products online and placing an order. If this insurance is purchased, the return shipping will be borne by the insurance company if the order is returned for any reason. Some online retailers on some platforms will provide consumers with return shipping insurance to improve service levels and increase sales.

Insurance companies that provide return shipping insurance as a third party need to conduct risk assessment and management to formulate pricing strategies for return shipping insurance, including premiums and compensation costs. At present, the pricing method of freight insurance provided by online platforms is relatively simple. The compensation fee is generally 8-12 RMB, and the insurance premium is simply determined according to the return rate of consumers in the previous period. Although the current return shipping insurance pricing method is effective to a certain extent, it does not fully consider the background and characteristics of e-commerce, it can neither meet the needs of consumers (especially their psychological needs), nor make insurance companies' Profit maximization.

SUMMARY OF THE INVENTION

The present invention provides a return shipping insurance pricing method, a pricing system and a construction method thereof, which are used to solve the technical problem that the existing return shipping insurance pricing cannot effectively take into account the needs of consumers and insurance profits.

The technical solution of the present invention to solve the above-mentioned technical problems is as follows: a method for pricing return shipping insurance, comprising:

Based on the customer's behavior data and purchase data of the currently purchased product, the return probability prediction model is used to predict the customer's return probability of the current product; the return probability is brought into the objective function with the goal of maximizing profit, and the return shipping insurance is calculated. price;

Wherein, the behavior data includes the data formed by the current commodity information and behavior information generated by the behaviors of clicking the commodity, collecting the commodity and adding to the shopping cart; the purchase data includes the order quantity and the order amount paid by the customer on the online platform , the number of returned orders, the returned item information, and the return amount.

The beneficial effects of the invention are as follows: the method is a pricing method based on a big data mining algorithm, which makes full use of the sufficient and large amount of customer behavior data and historical data owned by e-commerce enterprises, and evaluates the customer's return probability through a prediction model. , which not only meets the needs of consumers, but also maximizes the profits of insurance companies. It is a new and effective pricing method for return shipping insurance.

On the basis of the above technical solutions, the present invention can also be improved as follows.

Further, the construction method of the return probability prediction model is:

Use the behavior data and the purchase data of multiple customers for each commodity to construct a training set and a prediction set; and use the training set to train multiple models, and use the prediction set to select the model with the highest prediction accuracy, as Return probability prediction model;

The plurality of models include: Multilayer Perceptron, Random Forest, and Logistic Regression.

The further beneficial effects of the present invention are: pre-constructing a training set and a prediction set, training multiple models through the training set, and using the prediction set to select a model with better prediction performance as a return probability prediction model, thereby improving the objectivity of return shipping insurance pricing.

Further, before using the behavior data and the purchase data, the behavior data and the purchase data are cleaned, deduplicated, and devoid of preprocessing.

Further, the objective function is: π=pD(p, _r )-Pr C;

Among them, p represents the return freight insurance price of the product, C is the compensation cost, P _r is the return probability obtained by the return probability prediction model, D(p,r)=α-βp-γr represents when the return freight insurance price is p The customer's purchase intention when the product evaluation index is r, where α is the basic purchase probability, and β and γ are the sensitive indexes.

The present invention also provides a method for constructing a return shipping insurance pricing system, comprising:

Acquiring a plurality of data samples corresponding to the orders placed by the target customers, each data sample including the behavior data and purchase data described in the above-mentioned one return shipping insurance pricing method for the corresponding commodities by the target customers;

Based on each data sample, a return freight insurance pricing method as described above is used to obtain the return freight insurance price of the commodity corresponding to the data sample;

Each data sample and its corresponding return freight insurance price constitute a piece of training sample data to obtain a training sample set, and input the return probability prediction model described in the above-mentioned one return freight insurance pricing method, to return freight insurance The price is used as the prediction target, and the return freight insurance pricing model is obtained by training, which is used as the return freight insurance pricing system for the target customer.

The beneficial effects of the invention are: the optimal price obtained by using the above pricing method is the softest price target, the selected data mining model takes the return shipping insurance price as the prediction target, and the price predicted by the calculation model is between the price and the soft price target. The error is minimized, the error function is minimized, the model parameters are optimized, the final pricing system is obtained, and the construction efficiency of the pricing system is improved. At the same time, a pricing system is formulated for each customer, which is highly targeted and reasonable in pricing.

The present invention also provides a return shipping insurance pricing method, which acquires the behavior data and the purchase data of the current commodity of the target customer, inputs the return shipping insurance pricing system of the target customer obtained by the above method, and obtains the current product of the target customer. Return freight insurance price of the product, complete the return freight insurance pricing of the target customer's current product.

The beneficial effects of the present invention are: using the above-mentioned pricing system for each customer, the optimal pricing decision can be obtained in one step without manual participation, the system automatically provides the optimal return shipping insurance price for individual customers, and the calculation efficiency is high.

The present invention also provides a computer-readable storage medium, characterized in that the computer-readable storage medium includes a stored computer program, wherein when the computer program is run by a processor, the device where the storage medium is located is controlled to execute the above The method for pricing return freight insurance, the construction method for the above-mentioned method for pricing return freight insurance, and/or the above-mentioned method for pricing return freight insurance.

Description of drawings

1 is a flowchart of a method for pricing return shipping insurance provided by an embodiment of the present invention;

Fig. 2 is a flow chart of constructing a return shipping insurance pricing system provided by an embodiment of the present invention;

FIG. 3 is a flowchart of a specific implementation of the construction of a return shipping insurance pricing system provided by an embodiment of the present invention.

Detailed ways

In order to make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention, but not to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

Example 1

A pricing method for return shipping insurance, as shown in Figure 1, includes:

Based on the customer's behavior data and purchase data of the currently purchased product, the return probability prediction model is used to predict the customer's return probability of the current product; the return probability is brought into the objective function with the goal of maximizing profit, and the return freight insurance price is calculated.

Among them, the behavior data includes the data composed of the current product information and behavior information generated by the behavior of clicking the product, collecting the product and adding to the shopping cart; the purchase data includes the order quantity paid by the customer on the online platform, the order amount, and the return order. Quantity, returned item information, and return amount.

This method is a pricing method based on big data mining algorithm. This method makes full use of the sufficient and large amount of customer behavior data and historical data owned by e-commerce companies, and evaluates the customer's return probability through a prediction model, which not only meets the needs of consumers It can also maximize the profits of insurance companies, and it is a new and effective pricing method for return shipping insurance.

Preferably, the construction method of the above-mentioned return probability prediction model is:

Use the behavior data and purchase data of multiple customers for each commodity to construct a training set and a prediction set; and use the training set to train multiple models, and use the prediction set to select the model with the highest prediction accuracy as the return probability prediction model; The models include: Multilayer Perceptron, Random Forest, and Logistic Regression.

Build a training set and a prediction set in advance, train multiple models through the training set, and use the prediction set to select the model with better prediction performance as the return probability prediction model to improve the objectivity of return shipping insurance pricing.

Preferably, before the behavior data and the purchase data are used, the behavior data and the purchase data are cleaned, deduplicated, and empty preprocessed.

Preferably, the above objective function is: π=pD(p, _r )-Pr C;

Among them, p represents the return freight insurance price of the product, C is the compensation cost, P _r is the return probability obtained by the return probability prediction model, and D(p,r)=α-βp-γr represents when the return freight insurance price is p. The customer's purchase intention when the commodity evaluation index is r, where α is the basic purchase probability, and β and γ are the sensitive indexes. The commodity evaluation index here can be obtained through statistical analysis of commodity evaluation information, and the sensitivity index can be obtained through statistical analysis of customer historical data.

Embodiment 2

A construction method of a return shipping insurance pricing system, comprising:

Acquiring a plurality of data samples corresponding to the orders placed by the target customer, each data sample including the behavior data and purchase data described in the method for pricing return shipping insurance according to the first embodiment of the target customer for each corresponding commodity;

Based on each data sample, a return shipping insurance pricing method as described in Embodiment 1 is used to obtain the return shipping insurance price of the commodity corresponding to the data sample;

Each data sample and its corresponding return shipping insurance price constitute a piece of training sample data to obtain a training sample set, and input the return probability prediction model described in the method for pricing return shipping insurance as described in Embodiment 1, to obtain a training sample set. The return freight insurance price is used as the prediction target, and the return freight insurance pricing model is obtained by training, which is used as the return freight insurance pricing system for the target customer.

The optimal price obtained by using the pricing method of the first embodiment is the soft price target, the selected data mining model takes the return shipping insurance price as the prediction target, and the error between the price predicted by the model and the soft price target is the smallest. The error function is optimized to optimize the model parameters and obtain the final pricing system.

For example, FIG. 2 shows a schematic flowchart of the construction of a return shipping insurance pricing system based on big data technology proposed in this embodiment, as shown in FIG. 2 , including the following steps:

Step 1, customer behavior data and historical data extraction and preprocessing.

A large amount of customer behavior data and historical data are stored in the network logs of the online platforms of e-commerce enterprises, which are all data left by consumers' corresponding operations on the e-commerce website. These data are extracted from network logs for preprocessing, including data cleaning, deduplication, and emptying, and preliminary statistics to remove non-universal data and abnormal data.

Step 2: Predict the customer's return probability and build a pricing model to obtain the optimal price applicable to the customer, which is used as a soft target for training the pricing system.

Specifically, the upper part of FIG. 3 describes the flow chart of the specific implementation of this step. Taking the processed data as input, select a prediction model, which can be a data mining algorithm. Here, take the multi-layer perceptron prediction model as an example. The consumer's return probability, the predicted return probability is used for the price decision model, and the optimal price decision is obtained, which is called the price soft target, and will be used to guide the training of the pricing prediction model.

The objective function of the price decision model in this step is the profit function of each order of the insurance company, and the formula is as follows:

π=pD(p, _r )-Pr C

Among them, p represents the premium of this single insurance, C is the compensation cost, P _r is the customer return probability obtained in the prediction model of the previous step, D(p,r)=α-βp-γr represents when the return shipping insurance price is p , the customer's purchase intention when the product evaluation index is r, where α is the basic purchase probability, and β and γ are the sensitive indexes. Here we can set the customer's basic purchase probability α=0.5, the product evaluation index can be obtained by calculating the product's favorable rate, and the sensitivity index can be set as a value in the [0,1] interval through statistical analysis of the customer's historical data. numerical value.

Through this formula, taking profit maximization as the optimization goal, the optimal pricing is obtained, and it is used as the price soft objective to guide the training of the pricing system.

Step 3, using the obtained price soft target to train a price prediction model to build a pricing system.

Specifically, the lower part of FIG. 3 describes the flow chart of the specific implementation of this step. Take the prediction model selected in step 2, take the data obtained in step 1 as input, take the return shipping insurance price as the prediction template, and use the price soft target obtained in step 2 for model training, and calculate the price and price soft target predicted by the model. The cross-entropy between the two can minimize the error function, so as to optimize the model parameters and obtain the final pricing system, which can take the preprocessed customer behavior data and historical data as input, and then directly output the optimal pricing decision.

Therefore, this method is a pricing method based on big data mining algorithm, and proposes a one-step pricing system based on customer data. This method makes full use of the sufficient and large amount of customer behavior data and historical data owned by e-commerce companies, and evaluates the customer's return probability through a predictive model after preprocessing, and takes profit maximization as the goal, bringing the customer return probability value into the target. Function, calculate the optimal return freight insurance price, and finally take the obtained optimal price as the soft price target, guide the training selected prediction model to take the return freight insurance price as the prediction target, optimize the model parameters, and get the final pricing system. The embodiment of the present invention can effectively utilize customer behavior data and historical data, mine effective information from the data, optimize the return shipping insurance price, and build an accurate and effective pricing system.

Embodiment 3

A return freight insurance pricing method, obtaining the behavior data and purchase data of the current commodity of the target customer, inputting the return freight insurance pricing system of the target customer obtained by the method as described in the second embodiment, and obtaining the return of the current commodity of the target customer. Freight insurance price, complete the return freight insurance pricing of the target customer's current product.

The system of the second embodiment can obtain the optimal pricing decision in one step, without manual participation, and has high calculation efficiency, and the system automatically provides the optimal return shipping insurance price personalized for the customer.

The pricing method and system for return shipping insurance designed in the above embodiments can effectively use customer behavior data and historical data, mine effective information from the data, and convert it into optimal price information through a price decision model, so as to guide Training of price prediction models to build accurate and effective pricing systems. The pricing system constructed by the technical solution can take the preprocessed customer behavior data and historical data as input, so as to obtain the optimal return shipping insurance pricing for each individual customer, so as to maximize the profit of the insurance business. Different from the existing technology, the pricing method proposed in this paper considers the impact of the evaluation index of the product on the purchase of return shipping insurance, and mines the customer's potential return willingness information from big data, and formulates a personalized pricing plan for the customer. At the same time, the pricing system constructed in this paper can obtain the optimal pricing decision in one step, without manual participation, the system automatically gives the optimal return shipping insurance price for individual customers, and the calculation efficiency is high.

Embodiment 4

A computer-readable storage medium, characterized in that the computer-readable storage medium includes a stored computer program, wherein when the computer program is run by a processor, the device where the storage medium is located is controlled to execute as described in the first embodiment. The method for pricing the return freight insurance described in the above, the construction method for the return freight insurance pricing system as described in the second embodiment, and/or the return freight insurance pricing method as described in the third embodiment.

The related technical solutions are the same as those of the first embodiment to the third embodiment, and are not repeated here.

Those skilled in the art can easily understand that the above are only preferred embodiments of the present invention, and are not intended to limit the present invention. Any modifications, equivalent replacements and improvements made within the spirit and principles of the present invention, etc., All should be included within the protection scope of the present invention.

Claims (7)

1. A method for pricing return shipping insurance, characterized in that, comprising: Based on the customer's behavior data and purchase data of the currently purchased product, the return probability prediction model is used to predict the customer's return probability of the current product; the return probability is brought into the objective function with the goal of maximizing profit, and the return shipping insurance is calculated. price; Wherein, the behavior data includes the data formed by the current commodity information and behavior information generated by the behavior of clicking the commodity, collecting the commodity and adding to the shopping cart; the purchase data includes the order quantity and the order amount paid by the customer on the online platform , the number of returned orders, the returned item information, and the return amount. 2. a kind of return freight insurance pricing method according to claim 1, is characterized in that, the construction method of described return probability prediction model is: Use the behavior data and the purchase data of multiple customers for each commodity to construct a training set and a prediction set; and use the training set to train multiple models, and use the prediction set to select the model with the highest prediction accuracy, as Return probability prediction model; The plurality of models include: Multilayer Perceptron, Random Forest, and Logistic Regression. 3. A return shipping insurance pricing method according to claim 2, characterized in that, before using the behavior data and the purchase data, the behavior data and the purchase data are cleaned, deduplicated, Null preprocessing. 4. A kind of return freight insurance pricing method according to claim 1, is characterized in that, described objective function is: π=pD(p, _r )-PrC; Among them, p represents the return freight insurance price of the product, C is the compensation cost, P _r is the return probability obtained by the return probability prediction model, D(p,r)=α-βp-γr represents when the return freight insurance price is p The customer's purchase intention when the product evaluation index is r, where α is the basic purchase probability, and β and γ are the sensitive indexes. 5. A method for constructing a return shipping insurance pricing system, comprising: Obtain a plurality of data samples corresponding to the orders placed by the target customer, each data sample includes the behavior data described in the method for pricing return shipping insurance according to any one of claims 1 to 4 for each corresponding commodity by the target customer and purchase data; Based on each data sample, a return shipping insurance pricing method according to any one of claims 1 to 4 is used to obtain the return shipping insurance price of the commodity corresponding to the data sample; Each data sample and its corresponding return shipping insurance price constitute a piece of training sample data, a training sample set is obtained, and the return shipping insurance pricing method described in any one of claims 1 to 4 is input. The probability prediction model takes the return freight insurance price as the prediction target, and trains the return freight insurance pricing model as the return freight insurance pricing system for the target customer. 6. A return freight insurance pricing method, characterized in that the behavior data and the purchase data of the current commodity of the target customer are obtained, and the return freight insurance pricing system of the target customer obtained by the method as claimed in claim 5 is input. , get the return freight insurance price of the target customer's current product, and complete the return freight insurance pricing of the target customer's current product. 7. A computer-readable storage medium, characterized in that the computer-readable storage medium comprises a stored computer program, wherein, when the computer program is run by a processor, the device where the storage medium is located is controlled to execute as claimed in the claims A kind of return freight insurance pricing method according to any one of 1 to 4, construction method of a kind of return freight insurance pricing system as claimed in claim 5 and/or a kind of return freight insurance pricing as claimed in claim 6 method.

Images