A method, device and equipment for determining product transfer probability

Abstract

The embodiment of the present invention discloses a method, device and equipment for determining product transfer probability, the method comprising: obtaining an initial model for product docking, the initial model for product docking comprising at least one product to be docked and at least one corresponding target docking product; for each product to be docked, determining a target lag distance and a target transfer probability function between the product to be docked and the target docking product, the target transfer probability function being determined according to historical data; determining the transfer probability between the product to be docked and the corresponding target docking product according to the target lag distance and the target transfer probability function. The target transfer probability function is determined in advance through historical data, and the transfer probability is predicted through the target transfer probability function and the target lag distance. Since the target transfer probability function is determined through historical data, the time lag effect of the historical data is taken into account, so the accuracy of the determined transfer probability is high, further improving the accuracy of the prediction result.

Description

Technical Field

[0001] Embodiments of the present invention relate to the field of computer technology, and in particular, to a method, device and equipment for determining product transfer probability. Background Art

[0002] In statistical testing, the commonly used model is the Markov chain model. After assigning an initial value to the transition probability, the model needs to be optimized according to the corresponding constraints. The common approach is to update and optimize the transition probability based on the maximum entropy principle to maximize the cumulative state entropy of the model; the second is to automatically generate the model transition probability with the optimization goal of maximizing the economic loss of failure.

[0003] Existing technologies generally generate transition probabilities from an optimization perspective based on constraints. Although existing technologies can obtain the optimal transition probability distribution under constraints, they ignore the time lag effect of historical data, resulting in a low accuracy rate of the determined transition probability. Summary of the invention

[0004] The present invention provides a method, device and equipment for determining product transfer probability, so as to predict the transfer probability of item ownership.

[0005] In a first aspect, an embodiment of the present invention provides a method for determining a product transfer probability, the method comprising:

[0006] Acquire an initial product docking model, wherein the initial product docking model includes at least one product to be docked and at least one corresponding target docking product;

[0007] For each product to be docked, determining a target hysteresis distance and a target transition probability function between the product to be docked and a target docking product, wherein the target transition probability function is determined according to historical data;

[0008] The transfer probability between the to-be-connected product and the corresponding target connection product is determined according to the target hysteresis distance and the target transfer probability function.

[0009] In a second aspect, an embodiment of the present invention further provides a device for determining a product transfer probability, the device comprising:

[0010] An initial model acquisition module, used to acquire an initial model for product docking, wherein the initial model for product docking includes at least one product to be docked and at least one corresponding target docking product;

[0011] A function determination module, used to determine, for each product to be docked, a target hysteresis distance and a target transition probability function between the product to be docked and a target docking product, wherein the target transition probability function is determined according to historical data;

[0012] The probability determination module is used to determine the transition probability between the to-be-connected product and the corresponding target connection product according to the target hysteresis distance and the target transition probability function.

[0013] In a third aspect, an embodiment of the present invention further provides a computer device, the device comprising:

[0014] one or more processors;

[0015] a memory for storing one or more programs,

[0016] When the one or more programs are executed by the one or more processors, the one or more processors implement a product transfer probability determination method as described in any one of the embodiments of the present invention.

[0017] The embodiment of the present invention provides a method, device and equipment for determining product transfer probability, by obtaining an initial model of product docking, the initial model of product docking includes at least one product to be docked and at least one corresponding target docking product; for each product to be docked, a target lag distance and a target transfer probability function are determined, the target transfer probability function is determined according to historical data; the transfer probability between the product to be docked and the corresponding target docking product is determined according to the target lag distance and the target transfer probability function. The target transfer probability function is determined in advance through historical data, and then the target transfer probability function between the product to be docked and the target docking product is determined, and the transfer probability is predicted through the target transfer probability function and the target lag distance. Since the target transfer probability function is determined through historical data, the time lag effect of the historical data is taken into account, so the accuracy of the determined transfer probability is high, which further improves the accuracy of the prediction result. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 is a flow chart of a method for determining product transfer probability in Embodiment 1 of the present invention;

[0019] Figure 2 is a flow chart of a method for determining product transfer probability in Embodiment 2 of the present invention;

[0020] Figure 3 It is a structural schematic diagram of an initial model of product docking in the second embodiment of the present invention;

[0021] Figure 4a This is a result display diagram of predicting the transition probability of a self-transition type by using a composite transition probability formula and a lag distance in the second embodiment of the present invention;

[0022] Figure 4b This is a result display diagram of predicting the transition probability of an interactive transition type by using a composite transition probability formula and a lag distance in the second embodiment of the present invention;

[0023] Figure 5a This is an example diagram of implementing data fitting corresponding to a self-transfer type in Embodiment 2 of the present invention;

[0024] Figure 5b This is an example diagram of implementing data fitting corresponding to an interactive transfer type in Embodiment 2 of the present invention;

[0025] Figure 6 This is an example diagram of an implementation of a product docking process in Embodiment 2 of the present invention;

[0026] Figure 7 This is an example diagram of an implementation of determining a product docking path in Embodiment 2 of the present invention.

[0027] Figure 8 It is a structural schematic diagram of a device for determining product transfer probability in Embodiment 3 of the present invention;

[0028] Fig. 9 It is a structural diagram of a computer device in Embodiment 4 of the present invention. DETAILED DESCRIPTION

[0029] In order to make the purpose, technical solutions and advantages of the present application clearer, the embodiments of the present application will be further described in detail below in conjunction with the accompanying drawings. It should be clear that the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in the field without creative work are within the scope of protection of the present application.

[0030] Embodiment 1

[0031] Figure 1 A flowchart of a method for determining a product transfer probability provided in the first embodiment of the present application is provided. The method is applicable to the case of predicting the transfer probability of the ownership of an item. The method can be executed by a computer device, which can be composed of two or more physical entities or one physical entity. Generally speaking, the computer device can be a notebook, a desktop computer, a smart tablet, etc.

[0032] like Figure 1 As shown, a method for determining product transfer probability provided by this embodiment 1 specifically includes the following steps:

[0033] S101: Acquire an initial product docking model, where the initial product docking model includes at least one product to be docked and at least one corresponding target docking product.

[0034] In this embodiment, the product docking initial model can be specifically understood as a data model including different products and the docking relationship between products. The product to be docked can be specifically understood as a product whose ownership right is about to expire and is waiting for other products to dock with it; the target docking product can be specifically understood as a product that replaces the product to be docked after the ownership right of the product to be docked held by the user expires. The products involved in this application can be virtual products. The types of the product to be docked and the target docking product can be different, or products of the same type can be selected as the target docking product, or they can be the same product, that is, the target docking product of the product to be docked is still the product to be docked.

[0035] Specifically, different products are analyzed in advance, and an initial product docking model is generated and stored according to the docking relationship between different products. The storage method can be local storage or server storage. The initial product docking model is obtained from the corresponding data storage space.

[0036] S102: for each product to be docked, determine a target lag distance and a target transfer probability function between the product to be docked and the target docking product, wherein the target transfer probability function is determined according to historical data.

[0037] In this embodiment, the target lag distance can be specifically understood as the time interval between the time when the product ownership transfer occurs and the predicted time. For example, the product to be docked is A, and the target docking product is B. It is predicted that the probability that the user obtains the ownership of the target docking product B as a substitute for the product to be docked A after 1 year, where 1 year is the target lag distance. The target transfer probability function can be specifically understood as a function expression for calculating the transfer probability, and the transfer probability functions between different products to be docked and target docking products may be different.

[0038] Specifically, the product whose ownership right is about to expire in the initial model of product docking is taken as the product to be docked, and the probability of the product to be docked being replaced by the target docking product B after the target lag distance h period is predicted. Determine each product to be docked. After determining the product to be docked, since there may be one or more target docking products corresponding to each product to be docked, determine the target lag distance and target transfer probability function between the product to be docked and each corresponding target docking product in turn. The target lag distance can be set according to demand, or when performing probability prediction, the user can input it each time or in batches. Analyze and process the historical data, and obtain the target transfer probability function by fitting the distribution probability of the historical data. When performing probability prediction, the predetermined target transfer probability function can be directly obtained, or the historical data can be obtained, analyzed and processed in real time, and the target transfer probability function can be determined.

[0039] S103: Determine the transition probability between the to-be-connected product and the corresponding target connection product according to the target hysteresis distance and the target transition probability function.

[0040] The transfer probability is the probability that the user's ownership of the product A to be docked changes to the ownership of the target docking product B. The target transfer probability function is a function expression of the transfer probability y with respect to the variable target lag distance x. The transfer probability is calculated based on the target lag distance and the target transfer probability function to obtain the transfer probability between each product to be docked and the target docking product. Based on the transfer probability, the initial product docking model can be updated, and the product docking path can be generated, etc., to achieve effective prediction of the user's acquisition of product ownership.

[0041] It can be known that when docking the expiration of product ownership, the docking product selected by the customer depends not only on the current unexpired product, but also on the time the unexpired product is held. Based on inertial thinking, the more recently held products are, the more likely customers are to choose the same product for docking; considering the replacement of products, the earlier held products are, the more likely customers are to choose different docking products when the products expire. Therefore, the transfer probability is related to the target lag distance.

[0042] The embodiment of the present invention provides a method for determining product transfer probability, by obtaining an initial product docking model, the initial product docking model includes at least one product to be docked and at least one corresponding target docking product; for each product to be docked, a target lag distance and a target transfer probability function between the product to be docked and the target docking product are determined, and the target transfer probability function is determined according to historical data; the transfer probability between the product to be docked and each corresponding target docking product is determined according to the target lag distance and the target transfer probability function. The transfer probability function corresponding to each product to be docked is determined in advance through historical data, and then the target transfer probability function between the product to be docked and the target docking product is determined, and the transfer probability is predicted through the target transfer probability function and the target lag distance. Since the target transfer probability function is determined through historical data, the time lag effect of the historical data is considered, so the accuracy of the determined transfer probability is high, which further improves the accuracy of the prediction result.

[0043] Embodiment 2

[0044] Figure 2 This is a flow chart of a method for determining product transfer probability provided by Embodiment 2 of the present invention. The technical solution of this embodiment is further refined on the basis of the above technical solution, and specifically mainly includes the following steps:

[0045] S201. Obtain product connection requirement document.

[0046] In this embodiment, the product docking requirement document can be specifically understood as a document containing the docking relationship between each product. The product docking requirement document is formed and stored based on the relationship between each product in the actual application, and the storage space can be a local space or a server. The product docking requirement document is obtained from the corresponding storage space.

[0047] S202: Parse the product docking requirement document to determine at least one product to be docked and a target docking product corresponding to each product to be docked.

[0048] Specifically, the product docking requirement document is stored according to certain rules when storing data. Therefore, the product docking requirement document is parsed according to the storage rules to determine the products to be docked in the product docking requirement document and at least one target docking product corresponding to each product to be docked.

[0049] S203: Generate an initial product docking model according to each to-be-docking product and each corresponding target docking product.

[0050] Generate the product docking initial model based on the relationship between each product to be docked and the corresponding target docking products. The transition probability between products can be set or not in the product docking initial model. If the transition probability is set as the initial transition probability, it can be obtained by evenly distributing the probability of each edge (the relationship between products) according to the knowledge-free allocation method.

[0051] For example, Figure 3 A structural diagram of an initial product docking model is provided. Taking the initial probability set between each to-be-docking product and the target docking product as an example, the initial transfer probability is obtained by evenly distributing the probability of each edge according to the knowledge-free allocation method.

[0052] S204: Acquire an initial product docking model, where the initial product docking model includes at least one product to be docked and at least one corresponding target docking product.

[0053] S205: For each product to be docked, determine a target lag distance and a target transfer probability function between the product to be docked and the target docking product, wherein the target transfer probability function is determined according to historical data.

[0054] The target transfer probability function in the embodiment of the present application can be determined in real time, that is, it can be determined each time the transfer probability prediction is performed, or it can be predetermined. Assuming that the product ownership lasts for a long time, for example, the user has the ownership of the product A to be docked for 2 years, then the user needs to choose to obtain the ownership of the target docking product B after the ownership of the product A to be docked expires, that is, two years later. Therefore, historical data needs to be counted for a long period of time. On this basis, the historical data used when predicting the transfer probability at different times may be consistent. Therefore, in order to reduce the amount of calculation, the target transfer probability function can be predetermined and obtained during calculation. After the historical data is updated or at intervals, the target transfer probability function is recalculated to achieve the update of the target transfer probability function.

[0055] As an optional embodiment of this embodiment, this optional embodiment further optimizes the target transition probability function to be determined as A1-A4:

[0056] A1. Determine the historical transfer probability cloud map corresponding to the docking of the product to be docked and the target docking product.

[0057] In this embodiment, the historical transition probability cloud map can be specifically understood as a point cloud image formed by the transition probability and hysteresis distance obtained through historical data statistics. The historical data is obtained, and statistical analysis is performed on the historical data to obtain the transition probability and hysteresis distance when the to-be-connected product and the target connection product are connected, and the historical transition probability cloud map is generated.

[0058] As an optional embodiment of this embodiment, this optional embodiment further optimizes the historical transition probability cloud map corresponding to the docking between the to-be-docking product and the target docking product to A11-A13:

[0059] A11. Obtain the transfer frequency and total number of samples when the product to be docked and the target docking product are docked at different lag distances.

[0060] In this embodiment, the transfer frequency can be specifically understood as the number of users whose ownership rights of the products to be docked are changed to the ownership rights of the target docking products. For example, when the target docking product is B, the ownership rights of the products to be docked A are changed to the ownership rights of the products B. The total number of samples can be specifically understood as the number of users whose ownership rights of the products to be docked are changed to the ownership rights of other types of products. For example, the total number of products B, C, and D are changed from the products to be docked A. The historical data is statistically analyzed to determine the transfer frequency and total number of samples when the products to be docked and the target docking products are docked under different hysteresis distances. For example, the historical data contains information about each product (such as when it expires and what product is selected for docking after expiration). Based on this, the data is statistically analyzed. When |h|=1 year, the data of each day is analyzed to determine all the products to be docked that expire in one year, and further determine what type of products are selected for ownership docking after the product expires, and obtain the transfer frequency of selecting product x1 and the total number of samples of selecting products x1-xn; continue to analyze the next day, and so on, to obtain the transfer frequency and total number of samples when |h|=1. By analogy, we can get |h|=2, 3, ..., where |h| represents the size of the lag distance.

[0061] A12. Determine the historical transition probability based on the ratio of transition frequency to the total number of samples.

[0062] In this embodiment, the historical transfer probability can be specifically understood as the probability that the ownership of product A in the historical data is changed to the ownership of product B. The ratio of the transfer frequency and the total number of samples is calculated, and the ratio is the historical transfer probability.

[0063] A13. Generate a historical transition probability cloud map based on each historical transition probability and the corresponding lag distance.

[0064] Each historical transition probability has a corresponding lag distance, and a historical transition probability cloud diagram is drawn based on each historical transition probability and the corresponding lag distance.

[0065] A2. Determine the probability function type between the product to be docked and the target docking product. The probability function type includes a self-transfer type and an interactive transfer type.

[0066] In this embodiment, the probability function type can be specifically understood as the type of the function calculation formula of the transfer probability, and the self-transfer type is the transfer between product A and product A. The interactive transfer is the transfer between product B and product B.

[0067] Specifically, after the product to be docked and the target product to be docked are determined, their respective product types are also determined accordingly. The probability function type is determined according to the product types of the product to be docked and the target product to be docked. When the product types of the product to be docked and the target product to be docked are the same, the probability function type is a self-transfer type; when the product types of the product to be docked and the target product to be docked are different, the probability function type is an interactive transfer type.

[0068] A3. Screen from predetermined composite transition probability formulas according to the probability function type to determine a target transition probability formula.

[0069] In this embodiment, the composite transition probability formula can be specifically understood as a formula for calculating the transition probability, and the composite transition probability formula in this application is a formula of the logistic function type. The target transition probability formula can be specifically understood as a probability calculation formula corresponding to the probability function type between the product to be docked and the target docking product.

[0070] Specifically, a composite transfer probability formula is determined in advance according to the probability function type, and each composite transfer probability formula is associated and mapped with the probability function type. After determining the probability function type between the product to be docked and the target docking product, each composite transfer probability formula is screened to obtain a target transfer probability formula matching the probability function type.

[0071] As an optional embodiment of this embodiment, this optional embodiment further optimizes and includes B1-B4 before screening from the predetermined composite transition probability formula according to the probability function type:

[0072] B1. Determine the target energy function related to the lag distance according to the probability function type.

[0073] In this embodiment, the target energy function is an expression of an energy function related to the hysteresis distance, and the target energy function is a self-energy function or an interactive energy function. The self-energy function is an energy function when the product performs self-transfer; the interactive energy function is an energy function when the product performs interactive transfer. Different energy functions are pre-constructed, and the target energy function matching the energy function is screened out from the energy function according to the probability function type.

[0074] It can be known that the transition probability function is unidirectional and asymmetric. The transition probability function describes the functional relationship between the two-point transition probability and the lag distance, and reflects the graphical trend of the two-point transition probability changing with the lag distance. In time series analysis, the transition probability is often assumed to be second-order stationary or intrinsically stationary. After obtaining the experimental transition probability from the sample data, it is necessary to select a suitable composite transition probability formula to interpolate or fit it. Consider a group containing at most two elements, then the transition probability p ijThe general expression for (h) can be derived from the conditional probability of a Markov random field with K classes:

[0075]

[0076] Among them, U(i,j,h) is the energy function of the current node and the historical node with respect to the lag distance h, and its value is determined by product type i and product type j and the proximity relationship between them, and K is the total number of product types. It should be pointed out that the directionality of the transition probability is reflected by the direction of energy transfer in the energy function.

[0077] Since the selection of energy function or potential function is not unique, in order to construct a reasonable transition probability, this application defines the self-energy function as:

[0078]

[0079] For the case where j≠i corresponds to the interaction energy function, define:

[0080]

[0081] Among them, a represents the range parameter of the model, p i is the base value of the target docking product with product type i; p j is the base value of the target docking product with product type j, p i and p j Usually, it is set as the prior global type ratio of product type i and product type j respectively, and |h| represents the size of the hysteresis distance h. The hysteresis distance h is a variable with direction. This application only needs to use the value of h during calculation, so the modulus operation is performed on h.

[0082] B2. Determine the initial transfer probability formula based on the target energy function and the number of product categories.

[0083] In this embodiment, the number of product categories can be specifically understood as the total number of product types, that is, K in the above formula (1). The K value is determined according to the number of product categories, and formulas (2) and (3) are combined with formula (1) to obtain the expression of the initial transition probability formula.

[0084] Among them, the initial transition probability formula corresponding to the self-transition type can be expressed as:

[0085]

[0086] The initial transition probability formula corresponding to the interaction type can be expressed as:

[0087]

[0088] B3. Integrate the initial transition probability formula according to the hyperbolic tangent function expression to obtain the intermediate transition probability formula.

[0089] It can be known that the transition probability function is defined as a function of the transition probability with respect to the time lag h, which is a parameterized graphical representation. A valid transition probability function must have the following properties:

[0090] (1) Continuity. The transition probability function should be a continuous function of the transition probability with respect to the gradually increasing lag distance |h|.

[0091] (2) For the probability function of the self-transition type, it should satisfy p ii (0) = 1; for the probability function of the interactive transfer type, it should satisfy p ij (0)=0.

[0092] (3) When |h|→+∞, there is p ii (h) = p i ,p ij (h) = p j .

[0093] In order to meet the above three conditions, the present application integrates the hyperbolic tangent function into equations (4) and (5), and the intermediate transition probability formula corresponding to the self-transition type is finally obtained as follows:

[0094]

[0095] The intermediate transition probability formula corresponding to the interactive transition type is:

[0096]

[0097] Wherein, b in the formula is a hyperbolic parameter, and its function is to ensure that the calculated transition probability does not exceed the effective range.

[0098] B4. Process the intermediate transition probability formula according to the cosine function to obtain the composite transition probability formula corresponding to the probability function type.

[0099] Since a weak cavitation effect will occur when the probability is calculated using the intermediate transfer probability formula, the intermediate transfer probability formula is processed using a cosine function to obtain a composite transfer probability formula. Different probability function types correspond to different transfer probability formulas.

[0100] The weak hole effect refers to the gradually weakening peaks and troughs that appear alternately on the transfer probability function graph as the lag distance increases. When the ownership of two products is frequently docked, it often means that the model needs to consider the strong correlation between such products. These two adjacent types and their simultaneous occurrence can be called neighbor structures. This feature causes some experimental interaction transfer probabilities to peak at short lag distances. Due to the existence of the hole effect, the self-transition probability will have a trough value at a short lag distance. In order to reflect this phenomenon, the present application constructs a nested model by selecting a suitable cosine function to fit the weak hole effect that appears in the experimental transfer probability. Thus, the composite transfer probability formula corresponding to different probability types is obtained. The composite transfer probability formula in the present application is similar to the structure of the logistic transfer probability function.

[0101] When the probability function type is a self-transition type, the compound transition probability formula is:

[0102]

[0103] Among them, p ii (h) is the transition probability corresponding to the self-transfer type, i is the product type of the product to be docked and the target docking product, p i is the base value of the target docking product with product type i, a is the range parameter, b is the hyperbolic parameter, w is the wavelength parameter, and h is the hysteresis distance.

[0104] When the probability function type is the interactive transfer type, the compound transfer probability formula is:

[0105]

[0106] Among them, p ij (h) is the transfer probability corresponding to the interactive transfer type, i is the product type of the product to be connected, j is the product type of the target connection product, p j is the base value of the target docking product with product type j, a is the range parameter, b is the hyperbolic parameter, w is the wavelength parameter, and h is the hysteresis distance.

[0107] For example, Figure 4a The present invention provides a method for predicting the transition probability of a self-transition type by using a composite transition probability formula and a lag distance. For the composite transition probability formula of the self-transition type, the base value p i It is set to 0.4653, the range parameter a and wavelength w are selected as 16 and 20 respectively, and the value of the hyperbolic parameter b is given in the figure. By selecting different lag distances, the corresponding transition probability is predicted, and the resulting image is as follows Figure 4a shown. Figure 4bThe present invention provides a method for predicting the transition probability of interactive transition types by using a composite transition probability formula and a lag distance. For the composite transition probability formula of interactive transition types, the base value p j It is set to 0.3454, the range parameter a and wavelength w are selected as 25 and 15 respectively, and the value of the hyperbolic parameter b is given in the figure. The resulting image is as follows Figure 4b shown.

[0108] like Figure 4a and 4b As shown in the figure, the range parameter a is used to measure the distance of the fluctuation of the transfer probability curve, that is, the distance from the fluctuation of the curve to the stability; the base value p i and p j is the transition probability value after the curve tends to be stable.

[0109] A4. Perform data fitting based on the target transfer probability formula and the historical transfer probability cloud map to determine the target transfer probability function when the product to be docked and the target docking product are docked.

[0110] Specifically, the historical transition probability cloud map is a scatter plot of probability distribution. A suitable data fitting method, such as least squares estimation, is used to fit the target transition probability formula and the historical transition probability cloud map to obtain a target transition probability function. Data fitting by least squares estimation can improve the computational efficiency of the model. The target transition probability formula is a formula including multiple unknown parameters. The values ​​of each parameter in the target transition probability formula can be obtained by data fitting. The obtained target transition probability function only includes one variable, the lag distance, and is then used to predict the transition probability.

[0111] For example, Figure 5a This is an example diagram of data fitting implementation corresponding to a self-transition type provided by an embodiment of the present invention. The sample data in the figure constitute a historical transition probability cloud diagram. Figure 5a It can be clearly seen that with the increase of the lag distance, alternating peaks and trough values ​​appear in the historical transfer probability cloud diagram. Therefore, formula (8) is used for fitting to obtain the fitting curve as follows: Figure 5a As shown, the fitting curve is the target transfer probability function. Figure 5b This is an example diagram of data fitting implementation corresponding to an interactive transfer type provided by an embodiment of the present invention. Formula (9) is used for fitting to obtain a fitting curve such as Figure 5b The fitting curve is the target transfer probability function. In the data fitting process, the values ​​of the parameters in the formula are shown in Table 1:

[0112] Table 1 Estimated parameters in the target transition probability function

[0113]

[0114]

[0115] Among them, the residual sum of squares is used to indicate the data fitting effect. The smaller the residual sum of squares, the better the fitting effect. In practical applications, the threshold of the residual sum of squares can be set according to needs. When the residual sum of squares is greater than the threshold, the fitting effect is not good at this time. You can reselect the fitting method to fit until you get a target transfer probability function with better effect.

[0116] S206: Determine the transition probability between the to-be-connected product and the corresponding target connection product according to the target hysteresis distance and the target transition probability function.

[0117] The target lag distance is taken as h and then substituted into the corresponding target transfer probability function for calculation to obtain the transfer probability between the product to be docked and the corresponding target docking product.

[0118] S207: Update the initial product docking model according to each transition probability to obtain a target product docking model.

[0119] In this embodiment, the target product docking model can be specifically understood as a model including the docking relationship and transition probability between each product. Since the product docking initial model only includes the docking relationship between each product, the transition probability is not included, or even if the transition probability is included, the size of the transition probability is evenly distributed. Therefore, the transition probability between each product is updated according to the calculated transition probability.

[0120] S208. Determine a product docking path according to the target product docking model.

[0121] The product docking path is determined according to the transfer paths between the products in the target product docking model, for example, a path is randomly selected, or an optimal path is selected according to the size of the transfer probability.

[0122] As an optional embodiment of this embodiment, this optional embodiment further determines the product docking path to be optimized as C1-C4 according to the target product docking model:

[0123] C1. Select the target products to be connected from the target product connection model;

[0124] In this embodiment, the target product to be docked can be specifically understood as any product to be docked in the target product docking model. The target product docking model includes multiple products to be docked. However, when determining the product docking path, each product to be docked needs to be used as the target product to be docked in turn, or only one of the products to be docked needs to be selected as the target product to be docked. A product to be docked can be randomly selected from the target product docking model as the target product to be docked, or a product to be docked can be selected from the target product docking model as the target product to be docked according to a probability value.

[0125] As an optional embodiment of this embodiment, this optional embodiment further optimizes the process of selecting target products to be docked from the target product docking model as follows:

[0126] C11. Determine the occurrence probability of each product to be docked in the target product docking model.

[0127] In this embodiment, the occurrence probability can be specifically understood as the probability of the user selecting each product to be docked. The historical data is counted to calculate the occurrence probability of each product to be docked. For example, product A to be docked appears 6 times in total, and all products appear 10 times in total. Therefore, the occurrence probability of product A to be docked is 0.6. After determining the occurrence probability, execute step C12 or C13 to determine the target docking product.

[0128] C12. Determine the product to be docked corresponding to the maximum probability value of each occurrence probability as the target docking product; or,

[0129] Directly compare the occurrence probabilities of the products to be connected and determine the maximum probability value. Products to be connected with larger occurrence probability values ​​are products that users often choose. Therefore, the products to be connected corresponding to the maximum probability value are selected first and used as the target connection products.

[0130] C12. Generate a random number; determine the target occurrence probability corresponding to the random number, and determine the to-be-connected product corresponding to the target occurrence probability as the target connection product.

[0131] A random number is generated by a random number generator. In this application, since the probability values ​​are all less than 1, the range of the random number is selected within the interval of 0-1. If it is greater than 1, it is adjusted to a range less than or equal to 1 through normalization. Determine the probability of occurrence closest to this random number and use it as the target probability of occurrence. For example, the random number is 0.3, and the probability of occurrence is 0.1, 0.4, and 0.5 respectively. The probability of occurrence closest to 0.3 is 0.4. The probability of occurrence 0.4 is used as the target probability of occurrence, and the product to be docked corresponding to the target probability of occurrence is determined as the target docking product. If there is more than one probability of occurrence closest to the random number, you can randomly select one, or select according to a preset rule, for example, select a probability of occurrence with a larger value as the target probability of occurrence.

[0132] C2. Determine the transfer probability between the target product to be docked and the corresponding target docking products.

[0133] Specifically, the target docking product corresponding to the target product to be docked may be one or more target docking products, and the transfer probability between the target product to be docked and each corresponding target docking product is determined in sequence.

[0134] C3. Determine the maximum value among all transition probabilities.

[0135] Compare the transition probabilities and determine the maximum transition probability.

[0136] C4. Generate a product docking path according to the path between the target product to be docked and the target docking product corresponding to the maximum value.

[0137] The target docking product corresponding to the maximum value of the transfer probability is determined, and the path between the target docking product and the target product to be docked is used as the product docking path.

[0138] S209. Generate test cases according to the product docking path.

[0139] In actual application, it is necessary to simulate users obtaining ownership of different products to test system functions. The product docking path can reflect the probability of users obtaining ownership of different products. The product docking path is used as a parameter when generating test cases.

[0140] For example, Figure 6An implementation example diagram of a product docking process provided for an embodiment of the present application. Take the case where the products to be docked include two products to be docked and two target products to be docked as an example. The tester or staff logs in to the client before testing, and docks when it is determined that the product ownership has expired. Among them, both the product to be docked 31 and the product to be docked 32 can be docked with product ownership rights. The product to be docked 31 and the product to be docked 32 can respectively select the target docking product 33 or the target docking product 34 for docking. The product docking path can be determined based on the product docking relationship and probability, and the data can be submitted. After submitting the data, you can return to the login step and repeat the step of selecting the target docking product. Figure 6 The probability of executing the steps is included in , where a is the probability of executing the product ownership expiration docking after logging in to the client; probabilities b and c are the occurrence probabilities of products 1 and 2 to be docked, respectively; probabilities d and e are the transfer probabilities of the ownership of product 1 to be docked changing to target docking product 33 and target docking product 34; probabilities f and g are the transfer probabilities of the ownership of product 32 to be docked changing to target docking product 33 and target docking product 34; h and i are the probabilities of submitting data after determining the target docking product 33 or 34, respectively.

[0141] For example, Figure 7 This is an implementation example diagram for determining the product docking path. The tester logs in to the client, performs the product ownership expiration docking operation, selects the product to be docked 31 as the target product to be docked, compares the transfer probabilities of the target docking product 33 and the target docking product 34 corresponding to the product to be docked 31, selects the target docking product 34 with a larger probability value for docking, generates the product docking path and submits it to generate a test case.

[0142] The embodiment of the present invention provides a method for determining the product transfer probability, which sets the energy function in advance according to different transfer types, determines the target energy function by the probability function type, and integrates the target energy function to obtain a composite transfer probability formula, and eliminates the weak hole effect by fitting the weak hole effect, thereby improving the accuracy of the transfer probability. When predicting the docking of the product to be docked with the target docking product, the corresponding target lag distance of the target transfer probability function is determined, and then the transfer probability is predicted. Since the target transfer probability function is determined by historical data and the time lag effect of the historical data is taken into account, the accuracy of the determined transfer probability is relatively high, further improving the accuracy of the prediction result.

[0143] Embodiment 3

[0144] Figure 8 This is a schematic diagram of the structure of a device for determining product transfer probability provided in Embodiment 3 of the present invention. The device comprises: an initial model acquisition module 41, a function determination module 42 and a probability determination module 43.

[0145] Among them, the initial model acquisition module 41 is used to obtain the product docking initial model, and the product docking initial model includes at least one product to be docked and at least one corresponding target docking product; the function determination module 42 is used to determine the target lag distance and target transfer probability function between the product to be docked and the target docking product for each product to be docked, and the target transfer probability function is determined according to historical data; the probability determination module 43 is used to determine the transfer probability between the product to be docked and the corresponding target docking product according to the target lag distance and the target transfer probability function.

[0146] An embodiment of the present invention provides a device for determining product transfer probability, which determines a target transfer probability function in advance through historical data, and then determines a target transfer probability function between a to-be-connected product and a target connection product, and predicts the transfer probability through the target transfer probability function and the target lag distance. Since the target transfer probability function is determined through historical data and the time lag effect of the historical data is taken into account, the accuracy of the determined transfer probability is relatively high, further improving the accuracy of the prediction result.

[0147] Furthermore, the device also includes:

[0148] Document acquisition module, used to obtain product docking requirement documents;

[0149] A document parsing module, used to parse the product docking requirement document, determine at least one product to be docked, and a target docking product corresponding to each of the products to be docked;

[0150] The model generation module is used to generate an initial product docking model according to each of the products to be docked and the corresponding target docking products.

[0151] Furthermore, the function determination module 42 includes:

[0152] A cloud map determining unit, used to determine a historical transition probability cloud map corresponding to the docking of the to-be-docking product and the target docking product;

[0153] A probability type determination unit, used to determine the probability function type between the to-be-connected product and the target connection product, wherein the probability function type includes a self-transition type and an interactive transfer type;

[0154] A probability screening unit, used to screen from predetermined composite transition probability formulas according to the probability function type to determine a target transition probability formula;

[0155] The fitting unit is used to perform data fitting according to the target transfer probability formula and the historical transfer probability cloud map to determine the target transfer probability function when the product to be docked and the target docking product are docked.

[0156] Furthermore, the cloud map determination unit includes:

[0157] A data acquisition subunit is used to obtain the transfer frequency and total number of samples when the to-be-docked product and the target docking product are docked at different lag distances;

[0158] A historical probability determination subunit, used to determine the historical transition probability according to the ratio of the transition frequency to the total number of samples;

[0159] The cloud map generating subunit is used to generate a historical transition probability cloud map according to each of the historical transition probabilities and the corresponding lag distances.

[0160] Furthermore, the data acquisition subunit is specifically used to determine a target energy function related to the lag distance according to the probability function type before screening from the predetermined composite transfer probability formula according to the probability function type; determine an initial transfer probability formula according to the target energy function and the number of product categories; integrate the initial transfer probability formula according to the hyperbolic tangent function expression to obtain an intermediate transfer probability formula; and process the intermediate transfer probability formula according to the cosine function to obtain a composite transfer probability formula corresponding to the probability function type.

[0161] Furthermore, when the probability function type is a self-transition type, the compound transition probability formula is:

[0162]

[0163] Among them, p ii (h) is the transition probability corresponding to the self-transfer type, i is the product type of the product to be docked and the target docking product, p i is the base value of the target docking product with product type i, a is the range parameter, b is the hyperbolic parameter, w is the wavelength parameter, and h is the hysteresis distance.

[0164] When the probability function type is an interactive transfer type, the compound transfer probability formula is:

[0165]

[0166] Among them, p ij (h) is the transfer probability corresponding to the interactive transfer type, i is the product type of the product to be connected, j is the product type of the target connection product, p j is the base value of the target docking product with product type j, a is the range parameter, b is the hyperbolic parameter, w is the wavelength parameter, and h is the hysteresis distance.

[0167] Furthermore, the device also includes:

[0168] An updating module, used for updating the product docking initial model according to each of the transition probabilities to obtain a target product docking model;

[0169] A path determination module, used to determine a product docking path according to the target product docking model;

[0170] A test case generation module is used to generate test cases according to the product docking path.

[0171] Furthermore, the path determination module includes:

[0172] A product screening unit, used to screen target products to be docked from the target product docking model;

[0173] A probability determination unit, used to determine the transfer probability between the target product to be docked and each corresponding target docking product;

[0174] A maximum value determination unit, used to determine the maximum value among the transition probabilities;

[0175] A path generating unit is used to generate a product docking path according to a path between the target to-be-docking product and the target docking product corresponding to the maximum value.

[0176] Furthermore, the product screening unit is specifically used to determine the occurrence probability of each to-be-docked product in the target product docking model; determine the to-be-docked product corresponding to the maximum probability value of each of the occurrence probabilities as the target docking product; or, generate a random number, determine the target occurrence probability corresponding to the random number, and determine the to-be-docked product corresponding to the target occurrence probability as the target docking product.

[0177] The product transfer probability determination device provided in the embodiment of the present invention can execute the product transfer probability determination method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0178] Embodiment 4

[0179] Fig. 9 A schematic diagram of the structure of a computer device provided in Embodiment 4 of the present invention is shown in FIG. Fig. 9 As shown, the device includes a processor 50, a memory 51, an input device 52 and an output device 53; the number of processors 50 in the device can be one or more. Fig. 9 A processor 50 is taken as an example; the processor 50, memory 51, input device 52 and output device 53 in the device can be connected by a bus or other means. Fig. 9 The example of connecting through bus is taken in the following.

[0180] The memory 51 is a computer-readable storage medium that can be used to store software programs, computer executable programs and modules, such as program instructions / modules corresponding to the product transfer probability determination method in the embodiment of the present invention (for example, the initial model acquisition module 41, the function determination module 42 and the probability determination module 43 in the product transfer probability determination device). The processor 50 executes various functional applications and data processing of the device by running the software programs, instructions and modules stored in the memory 51, that is, realizing the above-mentioned product transfer probability determination method.

[0181] The memory 51 may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system and at least one application required for a function; the data storage area may store data created according to the use of the terminal, etc. In addition, the memory 51 may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other non-volatile solid-state storage device. In some instances, the memory 51 may further include a memory remotely arranged relative to the processor 50, and these remote memories may be connected to the device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0182] The input device 52 may be used to receive input digital or character information and generate key signal input related to user settings and function control of the device. The output device 53 may include a display device such as a display screen.

[0183] It is worth noting that in the embodiment of the above-mentioned product transfer probability determination device, the various units and modules included are only divided according to functional logic, but are not limited to the above-mentioned division, as long as the corresponding functions can be achieved; in addition, the specific names of the functional units are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of the present invention.

[0184] Note that the above are only preferred embodiments of the present invention and the technical principles used. Those skilled in the art will understand that the present invention is not limited to the specific embodiments described herein, and that various obvious changes, readjustments and substitutions can be made by those skilled in the art without departing from the scope of protection of the present invention. Therefore, although the present invention has been described in more detail through the above embodiments, the present invention is not limited to the above embodiments, and may include more other equivalent embodiments without departing from the concept of the present invention, and the scope of the present invention is determined by the scope of the appended claims.

Claims

1. A method for determining product transfer probability, characterized in that: include: Acquire an initial product docking model, wherein the initial product docking model includes at least one product to be docked and at least one corresponding target docking product; For each product to be docked, determining a target hysteresis distance and a target transition probability function between the product to be docked and a target docking product, wherein the target transition probability function is determined according to historical data; Determine the transfer probability between the to-be-connected product and the corresponding target connection product according to the target hysteresis distance and the target transfer probability function; Determine the target transition probability function, including: Determine the historical transition probability cloud map corresponding to the docking of the product to be docked and the target docking product; Determine a probability function type between the to-be-connected product and the target connection product, wherein the probability function type includes a self-transition type and an interactive transfer type; Screening from predetermined composite transition probability formulas according to the probability function type to determine a target transition probability formula; Data fitting is performed according to the target transfer probability formula and the historical transfer probability cloud map to determine the target transfer probability function when the product to be docked and the target docking product are docked.

2. The method according to claim 1, characterized in that Before obtaining the initial model for product docking, the method further includes: Obtain product docking requirements documents; Parsing the product docking requirement document to determine at least one product to be docked and a target docking product corresponding to each of the products to be docked; A product docking initial model is generated according to each of the to-be-docking products and each corresponding target docking product.

3. The method according to claim 1, characterized in that Determining the historical transition probability cloud map corresponding to the docking of the product to be docked and the target docking product, including: Obtain the transfer frequency and total number of samples when the product to be docked and the target docking product are docked at different lag distances; Determining the historical transition probability according to the ratio of the transition frequency to the total number of samples; A historical transition probability cloud diagram is generated according to each of the historical transition probabilities and the corresponding lag distances.

4. The method according to claim 1, characterized in that Before screening from the predetermined composite transition probability formula according to the probability function type, the method further includes: Determine the target energy function related to the lag distance according to the probability function type; Determine an initial transition probability formula based on the target energy function and the number of product categories; Integrate the initial transition probability formula according to the hyperbolic tangent function expression to obtain an intermediate transition probability formula; The intermediate transition probability formula is processed according to the cosine function to obtain a composite transition probability formula corresponding to the probability function type.

5. The method according to claim 4, characterized in that When the probability function type is a self-transition type, the compound transition probability formula is: Among them, p ii (h) is the transition probability corresponding to the self-transfer type, i is the product type of the product to be docked and the target docking product, p i is the base value of the target docking product with product type i, a is the range parameter, b is the hyperbolic parameter, w is the wavelength parameter, and h is the hysteresis distance; When the probability function type is an interactive transfer type, the compound transfer probability formula is: Among them, p ij (h) is the transfer probability corresponding to the interactive transfer type, i is the product type of the product to be connected, j is the product type of the target connection product, p j is the base value of the target docking product with product type j, a is the range parameter, b is the hyperbolic parameter, w is the wavelength parameter, and h is the hysteresis distance.

6. The method according to any one of claims 1 to 5, characterized in that: Also includes: The product docking initial model is updated according to each of the transition probabilities to obtain a target product docking model; Determining a product docking path according to the target product docking model; Generate test cases according to the product docking path.

7. The method according to claim 6, characterized in that The determining of the product docking path according to the target product docking model includes: Screening out target products to be docked from the target product docking model; Determine the transfer probability between the target product to be connected and each corresponding target connection product; Determining a maximum value among the transition probabilities; A product docking path is generated according to a path between the target to-be-docking product and the target docking product corresponding to the maximum value.

8. The method according to claim 7, characterized in that The step of selecting a target product to be docked from the target product docking model includes: Determine the occurrence probability of each product to be docked in the target product docking model; Determine the to-be-matched product corresponding to the maximum probability value of each of the occurrence probabilities as the target matching product; or, A random number is generated, a target occurrence probability corresponding to the random number is determined, and a to-be-connected product corresponding to the target occurrence probability is determined as a target connection product.

9. A device for determining product transfer probability, characterized in that: include: An initial model acquisition module, used to acquire an initial model for product docking, wherein the initial model for product docking includes at least one product to be docked and at least one corresponding target docking product; A function determination module, used to determine, for each product to be docked, a target hysteresis distance and a target transition probability function between the product to be docked and a target docking product, wherein the target transition probability function is determined according to historical data; A probability determination module, used to determine the transition probability between the to-be-connected product and the corresponding target connection product according to the target hysteresis distance and the target transition probability function; The function determination module comprises: A cloud map determining unit, used to determine a historical transition probability cloud map corresponding to the docking of the to-be-docking product and the target docking product; A probability type determination unit, used to determine the probability function type between the to-be-connected product and the target connection product, wherein the probability function type includes a self-transition type and an interactive transfer type; A probability screening unit, used to screen from predetermined composite transition probability formulas according to the probability function type to determine a target transition probability formula; The fitting unit is used to perform data fitting according to the target transfer probability formula and the historical transfer probability cloud map to determine the target transfer probability function when the product to be docked and the target docking product are docked.

10. A computer device, characterized in that: The device comprises: one or more processors; a memory for storing one or more programs, When the one or more programs are executed by the one or more processors, the one or more processors implement a product transfer probability determination method as described in any one of claims 1-8.

Images