Method, device and electronic device for analyzing the recommendation effect of prop recommendation link

Abstract

The present application discloses a method, device, electronic device, and computer-readable storage medium for analyzing the recommendation effect of a prop recommendation link. The method includes: obtaining a prop recommendation link, wherein the prop recommendation link includes each recommendation event including each intermediate event and each final recommendation result after the recommendation is completed; determining each first transition probability of each intermediate event directly transferring to each recommendation event; determining each second transition probability of each intermediate event transferring to each intermediate event based on each first transition probability; determining each third transition probability of each intermediate event transferring to each intermediate event and then directly transferring to each final recommendation result based on each second transition probability and the probability of each intermediate event directly transferring to each final recommendation result; and determining each probability of each intermediate event transferring to each final recommendation result based on each third transition probability. This solution can efficiently and conveniently analyze the prop recommendation effect, thereby efficiently improving the prop recommendation link.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to a method, device, electronic device and computer-readable storage medium for analyzing the recommendation effect of a prop recommendation link. Background Art

[0002] In virtual games, when players purchase items, they typically trigger a purchase request. Based on this request, the system finds the item that best meets the player's needs and recommends it to them. However, in real-world situations, the item that best meets the player's needs may not be recommended. For example, the item that best meets the player's needs may be sold out, or the player may not have enough tokens to purchase the item. Therefore, if the system detects that the item that best meets the player's needs cannot be successfully recommended, it will often recommend an alternative item.

[0003] In actual applications, it's common for players to refuse to purchase certain alternative items when the system recommends them. Alternatively, alternative items may not even be recommended to players. These phenomena often indicate that the item recommendation chain set by the developer may be inappropriate. Developers need to modify the inappropriate item recommendation chain and analyze the recommendation results of the modified item recommendation chain to determine the optimal item recommendation chain.

[0004] In related technologies, developers can split the item recommendation chain into complete item recommendation branches, calculate the recommendation effect corresponding to each item recommendation branch, and modify the item recommendation chain based on the recommendation effect. However, due to the wide variety of alternative items in the item recommendation chain, when an alternative item is unsuccessful, the alternative item has its own alternatives. Therefore, the item recommendation chain is divided into a wide variety of item recommendation branches and is extremely complex. Using the above solution to calculate the results of each link will require a large number of complex and tedious calculations, making the analysis of the recommendation effect of the item recommendation chain less efficient. Summary of the Invention

[0005] This application provides a method, device, electronic device, and computer-readable storage medium for analyzing the recommendation effect of a prop recommendation link, which can more conveniently and efficiently analyze the recommendation effect of a prop recommendation link. The specific solution is as follows.

[0006] In a first aspect, an embodiment of the present application provides a method for analyzing the recommendation effect of a prop recommendation link, the method comprising:

[0007] Obtaining a prop recommendation link, wherein the prop recommendation link includes various recommendation events, including intermediate events and final recommendation results after the recommendation is completed. The intermediate events are used to represent the intermediate situations between the start and end of the recommendation;

[0008] For each of the intermediate events, determining first transition probabilities of the intermediate event directly transitioning to each of the recommended events;

[0009] For each of the intermediate events, calculating, based on each of the first transition probabilities, a probability of the intermediate event transitioning to each of the final recommendation results;

[0010] The calculating, according to each first transition probability, the probability of the intermediate event being transferred to each final recommendation result includes:

[0011] Determining second transition probabilities of transferring the targeted intermediate event to the intermediate events according to the first transition probabilities;

[0012] Determine each third transition probability based on each second transition probability and each target transition probability; the target transition probability is each first transition probability of each intermediate event directly transitioning to each final recommendation result in each first transition probability, and the third transition probability is the probability that the targeted intermediate event transitions to each intermediate event and then directly transitions from the transferred intermediate event to each final recommendation result;

[0013] The probability of the intermediate event transitioning to each of the final recommendation results is determined according to each of the third transition probabilities.

[0014] In a second aspect, an embodiment of the present application provides a device for analyzing the recommendation effect of a prop recommendation link, the device comprising: an acquisition unit for acquiring a prop recommendation link, the prop recommendation link including recommendation events, the recommendation events including intermediate events and final recommendation results after the recommendation is completed, the intermediate events being used to represent intermediate situations occurring between the start of the recommendation and the end of the recommendation;

[0015] a determining unit, configured to determine, for each of the intermediate events, first transition probabilities of the intermediate event directly transitioning to the recommended events;

[0016] a calculation unit, configured to calculate, for each of the intermediate events, a probability of the intermediate event transitioning to each of the final recommendation results according to each of the first transition probabilities;

[0017] The calculation unit is specifically used to: determine the second transition probabilities of the intermediate event being targeted transferring to each of the intermediate events based on the first transition probabilities; determine the third transition probabilities based on the second transition probabilities and the target transition probabilities; the target transition probabilities are the first transition probabilities of each of the intermediate events being directly transferred to each of the final recommendation results in the first transition probabilities, and the third transition probabilities are the probabilities of the intermediate event being targeted transferring to each of the intermediate events and then being directly transferred from the transferred intermediate event to each of the final recommendation results; determine the probability of the intermediate event being targeted transferring to each of the final recommendation results based on the third transition probabilities.

[0018] In a third aspect, the present application further provides an electronic device, comprising:

[0019] processor; and

[0020] The memory is used to store a data processing program. After the electronic device is powered on and the program is run by the processor, the method as described in any one of the first aspects is executed.

[0021] In a fourth aspect, an embodiment of the present application further provides a computer-readable storage medium storing a data processing program, which is run by a processor to execute the method as described in any one of the first aspects.

[0022] Compared with the prior art, this application has the following advantages:

[0023] The analysis method of the recommendation effect of the prop recommendation link provided in the present application can obtain the prop recommendation link, which includes various recommendation events. Each recommendation event includes: various intermediate events and various final recommendation results after the recommendation is completed. The intermediate events are used to represent the intermediate situations that occur between the start of the recommendation and the end of the recommendation; for each intermediate event, determine the first transition probabilities of the intermediate event directly transferring to each recommended event; determine the second transition probabilities of the intermediate event transferring to each intermediate event based on the first transition probabilities; determine the third transition probabilities based on the second transition probabilities and the target transfer probabilities; the target transfer probability is the first transition probability of each intermediate event directly transferring to each final recommendation result in each first transition probability, and the third transition probability is the probability that the intermediate event transfers to each intermediate event and the intermediate event transferred to is directly transferred to each final recommendation result; determine the probability of the intermediate event transferring to each final recommendation result based on the third transition probability. Since the intermediate event can be directly transferred to the final recommendation result, or first transferred to the intermediate event and then transferred to the final recommendation result from the intermediate event, when calculating the probability of the intermediate event transferring to the final recommendation result, you can first calculate the second transfer probability of the intermediate event transferring to each intermediate event, and then calculate the probability of the intermediate event transferring to the final recommendation result based on the probability of each intermediate event directly transferring to the final recommendation result.

[0024] It can be seen that the analysis method of the recommendation effect of the prop recommendation link provided in the present application splits the probability of each intermediate event in the prop recommendation link being transferred to each final recommendation result into the probability of each intermediate event being transferred to the intermediate event and the probability of each intermediate event being directly transferred to each final recommendation result. This makes it easier to calculate the probability of each intermediate event in the prop recommendation link being transferred to each final recommendation result, and the corresponding recommendation effect statistics are more efficient, which simplifies the calculation process, improves the calculation efficiency, and thereby improves the analysis efficiency of the recommendation effect of the prop recommendation link.

[0025] In addition, the analysis method of the recommendation effect of the prop recommendation link provided in this application uses a method with clear logic and good visualization effect to show the process of each intermediate state finally transferring to each final recommendation result, so that users can quickly understand the recommendation method of the prop recommendation link. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 This is a flow chart of a method for analyzing the recommendation effect of a prop recommendation link provided in an embodiment of the present application;

[0027] Figure 2 This is an example of a recommendation flow chart of a prop recommendation link provided in an embodiment of the present application;

[0028] Figure 3This is another example of a recommendation flow chart of the prop recommendation link provided in an embodiment of the present application;

[0029] Figure 4 is a transition matrix diagram determined by each first transition probability in the historical data provided by the embodiment of the present application;

[0030] Figure 5 This is a structural block diagram of a device for analyzing the recommendation effect of a prop recommendation link provided in an embodiment of the present application;

[0031] Figure 6 This is a structural block diagram of an example of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0032] The following description sets forth many specific details to facilitate a thorough understanding of the present application. However, the present application can be implemented in many other ways than those described herein, and those skilled in the art can make similar generalizations without violating the scope of the present application. Therefore, the present application is not limited to the specific implementations disclosed below.

[0033] It should be noted that the terms "first", "second", "third", etc. in the claims, description and drawings of the present application are used to distinguish similar objects and are not used to describe a specific order or sequence. The data used in this way are interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in a sequence other than that illustrated or described herein. In addition, the terms "including", "having" and their variations are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0034] In virtual games, when players purchase items, they typically trigger a purchase request. Based on this request, the system finds the item that best meets the player's needs and recommends it to them. However, in real-world situations, the item that best meets the player's needs may not be recommended. For example, the item that best meets the player's needs may be sold out, or the player may not have enough tokens to purchase the item. Therefore, if the system detects that the item that best meets the player's needs cannot be successfully recommended, it will often recommend an alternative item.

[0035] In actual applications, it's common for players to refuse to purchase certain alternative items when the system recommends them. Alternatively, alternative items may not even be recommended to players. These phenomena often indicate that the item recommendation chain set by the developer may be inappropriate. Developers need to modify the inappropriate item recommendation chain and analyze the recommendation results of the modified item recommendation chain to determine the optimal item recommendation chain.

[0036] In related technologies, developers can split the item recommendation chain into complete item recommendation branches, calculate the recommendation effect corresponding to each item recommendation branch, and modify the item recommendation chain based on the recommendation effect. However, due to the wide variety of alternative items in the item recommendation chain, when an alternative item is unsuccessful, the alternative item has its own alternatives. Therefore, the item recommendation chain is divided into a wide variety of item recommendation branches and is extremely complex. Using the above solution to calculate the results of each link will require a large number of complex and tedious calculations, making the analysis of the recommendation effect of the item recommendation chain less efficient.

[0037] Based on the above reasons, in order to analyze the recommendation effect of the prop recommendation link more conveniently and efficiently, the first embodiment of the present application provides an analysis method for the prop recommendation link, which is applied to electronic devices. The electronic devices can be desktop computers, laptops, mobile phones, tablets, servers, terminal devices, etc., or other electronic devices that can perform data statistics. The embodiments of the present application are not specifically limited.

[0038] Next, combine Figure 1 , the analysis method of the prop recommendation link recommendation effect provided in the embodiment of the present application is described in detail, Figure 1 This is a flowchart of the method for analyzing the recommendation effect of the prop recommendation link provided in an embodiment of the present application.

[0039] like Figure 1 As shown, the analysis method for the recommendation effect of the prop recommendation link provided in the embodiment of the present application includes Figure (a) and Figure (b), wherein Figure (a) includes steps 101 to 103, and Figure (b) includes steps 103-1 to 103-3, and step 103 in Figure (a) is implemented according to steps 103-1 to 103-3 in Figure (b).

[0040] Step 101: Get the item recommendation link.

[0041] Among them, the prop recommendation link includes various recommendation events, and each recommendation event includes: various intermediate events and various final recommendation results after the recommendation is completed. The intermediate events are used to represent the intermediate situations that occur between the start and end of the recommendation.

[0042] It is understandable that in a virtual game, the item recommendation instruction can be actively triggered by the user selecting the corresponding control, or it can be passively triggered by the user upgrading or the teammate upgrading. After receiving the item recommendation instruction, the server can recommend items to the user. Among them, the item recommendation instruction may include: the purpose of the item, the number of tokens held by the user, so that the server can find the target item that best meets the user's needs according to the item recommendation instruction; the target item can also be specified by the user in the item recommendation instruction, so that the item recommendation instruction can also include the item identifier of the target item, and the item identifier can be the item name or item number of the target item preset by the server, and the item identifier and the item are in a one-to-one correspondence. The triggering of the item recommendation instruction and the content contained in the item recommendation instruction are not specifically limited in this embodiment of the application.

[0043] In a specific implementation, the target item may not be successfully launched, for example, the number of game tokens held by the user is insufficient to purchase the target item, or the number of times the user has used the target item within the preset time has exceeded the preset number of uses (for example, if the server has preset a preset number of uses of target item A of 3 times, then if the user has already purchased target item A 3 times in May, then the user will not be able to purchase target item A again in May). Therefore, the developer can preset an item recommendation link in the server. The item recommendation link can be understood as item recommendation rules / methods / methods / mechanisms / logic, etc., to avoid launching other items to the user to the greatest extent when the target item cannot be successfully launched.

[0044] like Figure 2 As shown, it is a recommendation flow chart of an example of a prop recommendation link provided by an embodiment of the present application. Figure 2 The prop recommendation link shown in the example is: after receiving the prop recommendation instruction triggered by the user end, look for the target prop that best meets the user's needs, and when the target prop is found, determine whether the target prop can be pushed out, and when the target prop can be pushed out, recommend the target prop to the user; when the target prop is not found and / or the target prop cannot be pushed out, look for alternative props, determine whether the alternative prop can be pushed out, and when the alternative prop can be pushed out, recommend the alternative prop to the user; when no alternative prop is found and / or the alternative prop cannot be pushed out, look for popular props, determine whether the popular prop can be pushed out, and when the popular prop can be pushed out, recommend the popular prop to the user; when no popular prop is found and / or the popular prop cannot be pushed out, look for guaranteed props, determine whether the guaranteed prop can be pushed out, and when the guaranteed prop can be pushed out, recommend the guaranteed prop to the user.

[0045] It can be understood that the recommendation order of each prop in the prop recommendation link is determined according to the similarity of the prop information between the recommended prop and the target prop. The prop information may include but is not limited to the prop type, prop purpose, prop function, and tokens required to purchase the prop. The prop type may include but is not limited to attack props, defense props, auxiliary props, etc. This application does not specifically limit this.

[0046] The target item can be determined by the server based on the item recommendation command, or it can be a user-specified item. The alternative item can be the item with the highest item information similarity to the target item. This way, if the target item that best meets a user's needs cannot be recommended, using the alternative item can largely satisfy the user's needs, increasing the probability that the user will accept the recommended alternative item.

[0047] Popular items can be items that have been purchased more than a preset number of times within a preset timeframe and have the highest item information similarity to the target item. This allows users to more readily accept popular items with a higher purchase frequency when the target item that best meets their needs and the alternative item with the highest similarity to the target item are unavailable.

[0048] The guaranteed item can be the guaranteed item corresponding to the target item preset by the server. When the target item, alternative item, and popular item cannot be recommended to the user, the guaranteed item corresponding to the target item can be used to recommend it to the user. In actual applications, the server can set a guaranteed item for each type of item. The guaranteed item is often the item with the lowest purchase conditions among the items of that type. The purchase conditions include but are not limited to the level of the virtual character required for purchase, the tokens required for purchase, etc.

[0049] It should be noted that Figure 2 This is only an example of a recommendation process for the prop recommendation link provided in the embodiment of the present application. This recommendation process is provided to facilitate understanding of the present application and is not intended to limit the present application.

[0050] In practice, the recommendation events in an item recommendation chain include multiple intermediate events and at least one final recommendation result. An intermediate event is an intermediate state where an item recommendation can transition to other recommendation events (including both intermediate events and final recommendation results). A final recommendation result is a final state where an item recommendation cannot transition to other recommendation events. In other words, in an item recommendation chain, starting from receiving an item recommendation instruction and searching for a target item, various intermediate events may occur before ultimately reaching one or more final recommendation results.

[0051] Through this technical means, the complex prop recommendation chain is transformed into clear and concise intermediate events and final recommendation results, providing a basis for the analysis of the recommendation effect of the prop recommendation chain.

[0052] Step 102: For each intermediate event, determine first transition probabilities of the intermediate event directly transitioning to each recommended event.

[0053] It can be understood that in this step, the first transfer probability of the intermediate event directly transferring to the recommended events is the probability of transferring to each recommended event when the intermediate event undergoes one-step transfer. In the prop recommendation link, an intermediate event can be directly transferred to a final recommendation result or indirectly transferred to a final recommendation result. Indirect transfer means that the intermediate event is first transferred to each intermediate event, and then directly transferred to a final recommendation result by each intermediate event.

[0054] The following two methods are provided to obtain the first transition probabilities of each intermediate event directly transferring to each recommended event. Method 1: The developer collects all historical data within a preset time period, and counts and calculates the first transition probabilities of each intermediate event directly transferring to each recommended event from the collected historical data. The historical data can be obtained from the game logs recorded by the server. Method 2: The developer counts part of the historical data within a preset time period, and predicts other data based on part of the data. For example, if the number of recommended events is 10, the probability of a certain intermediate event directly transferring to 5 of the recommended events is counted, and then the probability of the intermediate event directly transferring to another 5 recommended events is predicted. Method 1 can accurately obtain each first transition probability based on historical data, and Method 2 can quickly and efficiently obtain each first transition probability when the amount of data is large.

[0055] It can be understood that, since in the item recommendation link, the intermediate event corresponds to the event that occurs in the next step, therefore, for each intermediate event, the sum of the first transition probabilities of the intermediate event directly transferring to the recommended events is 1.

[0056] Step 103: For each intermediate event, calculate the probability of the intermediate event transitioning to each final recommendation result according to each first transition probability.

[0057] It can be understood that the probability of the targeted intermediate event being transferred to each final recommendation result is the probability that the targeted intermediate event is finally transferred to each final recommendation result, and the probability that the targeted intermediate event is finally transferred to each final recommendation result includes the targeted intermediate event being directly transferred to each final recommendation result and the targeted intermediate event being indirectly transferred to each final recommendation result. That is, the process of transferring from the targeted intermediate event to the final recommendation result can be that the targeted intermediate event is transferred to any intermediate event and then transferred to the final recommendation result, or the targeted intermediate event is directly transferred to the final recommendation result. Therefore, the probability of each intermediate event being transferred to each final recommendation result can be indirectly calculated based on the first transfer probability of each intermediate event directly transferring to each recommended event.

[0058] Step 103 can be implemented according to the following steps 103-1 to 103-3.

[0059] Step 103 - 1 : Determine second transition probabilities of transitioning from the intermediate event to the intermediate events according to the first transition probabilities.

[0060] It is understandable that for each intermediate event, before transferring to the final recommendation result, the intermediate event can be transferred to each intermediate event first, and then directly transferred from each intermediate event to the final recommendation result. The intermediate event can be transferred to each intermediate event directly or indirectly.

[0061] Therefore, in this step, the second transition probability of each intermediate event transferring to each intermediate event is calculated, and the second transition probability is the sum of the first transition probabilities of the intermediate event directly transferring to each intermediate event and the probability of the intermediate event indirectly transferring to each intermediate event.

[0062] By using this technical means, the second transfer probabilities of the intermediate event being transferred to the intermediate events are obtained as the maximum probabilities that the intermediate event can be transferred to the intermediate events.

[0063] Step 103 - 2 : Determine each third transition probability according to each second transition probability and each target transition probability.

[0064] Among them, the target transfer probability is the first transfer probability of each intermediate event in each first transfer probability directly transferring to each final recommendation result, and the third transfer probability is the probability that the targeted intermediate event transfers to each intermediate event and then directly transfers to each final recommendation result from the transferred intermediate event.

[0065] In step 103-1, the maximum probability (second transition probability) that the intermediate event being targeted can be transferred to each intermediate event is obtained. To calculate the probability that the intermediate event being targeted will ultimately transfer to each final recommendation result, the probability of each intermediate event directly transferring to each final recommendation result after the intermediate event being targeted is transferred to each intermediate event can be calculated. Therefore, in step 103-2, the third transition probabilities of the intermediate event being transferred to each intermediate event first and then transferring to the final recommendation result are calculated. That is, the third transition probabilities can be further calculated based on the second transition probability and the first transition probabilities of each intermediate event directly transferring to each final recommendation result.

[0066] Step 103 - 3 : Determine the probability of the intermediate event transitioning to each of the final recommendation results according to each of the third transition probabilities.

[0067] As you can understand, there are multiple intermediate events in the item recommendation chain. Therefore, for each intermediate event, there are multiple calculated third recommendation probabilities, where the intermediate event transfers to each intermediate event and then directly transfers to the final recommendation result. The total number of third recommendation probabilities is the same as the total number of intermediate events. Each third transfer probability represents the probability that the intermediate event transfers to an intermediate event and then to the final recommendation result.

[0068] Therefore, in this step, the probability of the intermediate event being transferred to each of the final recommendation results can be calculated based on each third transfer probability corresponding to the intermediate event.

[0069] The analysis method of the recommendation effect of the prop recommendation link provided in the present application can obtain the prop recommendation link, which includes various recommendation events. Each recommendation event includes: various intermediate events and various final recommendation results after the recommendation is completed. The intermediate events are used to represent the intermediate situations that occur between the start of the recommendation and the end of the recommendation; for each intermediate event, determine the first transition probabilities of the intermediate event directly transferring to each recommended event; determine the second transition probabilities of the intermediate event transferring to each intermediate event based on the first transition probabilities; determine the third transition probabilities based on the second transition probabilities and the target transfer probabilities; the target transfer probability is the first transition probability of each intermediate event directly transferring to each final recommendation result in each first transition probability, and the third transition probability is the probability that the intermediate event transfers to each intermediate event and the intermediate event transferred to is directly transferred to each final recommendation result; determine the probability of the intermediate event transferring to each final recommendation result based on the third transition probability. Since the intermediate event can be directly transferred to the final recommendation result, or first transferred to the intermediate event and then transferred to the final recommendation result from the intermediate event, when calculating the probability of the intermediate event transferring to the final recommendation result, you can first calculate the second transfer probability of the intermediate event transferring to each intermediate event, and then calculate the probability of the intermediate event transferring to the final recommendation result based on the probability of each intermediate event directly transferring to the final recommendation result.

[0070] It can be seen that the analysis method of the recommendation effect of the prop recommendation link provided in the present application splits the probability of each intermediate event in the prop recommendation link being transferred to each final recommendation result into the probability of each intermediate event being transferred to the intermediate event and the probability of each intermediate event being directly transferred to each final recommendation result. This makes it easier to calculate the probability of each intermediate event in the prop recommendation link being transferred to each final recommendation result, and the corresponding recommendation effect statistics are more efficient, which simplifies the calculation process, improves the calculation efficiency, and thereby improves the analysis efficiency of the recommendation effect of the prop recommendation link.

[0071] In addition, the analysis method of the recommendation effect of the prop recommendation link provided in this application uses a method with clear logic and good visualization effect to show the process of each intermediate state finally transferring to each final recommendation result, so that users can quickly understand the recommendation method of the prop recommendation link.

[0072] Optional, such as Figure 3 FIG. 1 is another recommended flow chart provided in an embodiment of the present application. Figure 3The item recommendation link shown in the example is: after receiving the item recommendation instruction triggered by the user end, find the target item that best meets the user's needs; when the target item is found and recommended to the user, one-step transfer can occur to the following two situations: user acceptance, user rejection; when the target item is not found, one-step transfer can occur to the following eight situations: selecting associated alternative items, selecting popular alternative items, alternative items do not exist, alternative items exceed the user limit and cannot be launched, alternative items cannot be launched in the current scenario, alternative items cannot be launched except for exceeding the user limit and cannot be launched in the current scenario, and recommending guaranteed items; when alternative items do not exist, alternative items exceed the user limit and cannot be launched, alternative items cannot be launched in the current scenario, and alternative items cannot be launched except for exceeding the user limit and cannot be launched in the current scenario, one-step transfer can occur to the following two situations: selecting guaranteed items and items not launched; when selecting associated alternative items, selecting popular alternative items, and selecting guaranteed items, one-step transfer can occur to the following two situations: user acceptance and user rejection. In the subsequent steps of the embodiment of the present application, Figure 3 The following describes the example prop recommendation link.

[0073] Correspondingly, each intermediate event in step 101 may include at least one of the following:

[0074] Target item found, target item not found, related alternative items recommended, popular alternative items recommended, alternative items do not exist, alternative items cannot be launched due to exceeding user quota, alternative items cannot be launched in the current scenario, alternative items cannot be launched except for exceeding user quota and in the current scenario, and guaranteed items recommended.

[0075] The final recommendation results in step 101 may include the following results: user acceptance, user rejection, and no item is released.

[0076] Among them, the application scenario in which the substitute props cannot be launched due to exceeding the user limit may be: the server has preset a maximum number of purchases of the substitute props within a preset time period. If it is detected that the number of times the user has purchased the substitute props within the preset time period has reached the maximum number of purchases, the substitute props cannot be pushed to the user; the application scenario in which the substitute props cannot be launched in the current scenario may be: when the user sends virtual gifts to other users or to NPCs (non-player characters) in the game, equipment weapons are not applicable in this situation and therefore cannot be launched; the application scenario in which the substitute props cannot be launched for other reasons other than exceeding the user limit and being unable to be launched in the current scenario may be: the user's level has not reached the preset minimum usage level of the substitute props, the number of game tokens held by the user is insufficient to purchase the substitute props, etc.

[0077] For ease of statistics and viewing, in practical applications, each recommended event can be represented by an event identifier. Table 1 shows an example table of recommended events and event identifiers provided by an embodiment of the present application. S1 to S9 are intermediate events, and S10 to S12 are final recommendation results.

[0078] Table 1. Example table of recommended events and event identifiers

[0079]

[0080] As shown in Table 2, it is a table of first transition probabilities of each intermediate event transferring to a recommended event provided by an embodiment of the present application, wherein P x1,y1 represents the first transition probability from the intermediate event s-x1 directly to the recommended event s-y1, such as P 1,1 The probability of transferring directly from intermediate event s1 to intermediate event s2 is P 1,1 , P 1,12 The probability of transferring directly from the intermediate event s1 to the final recommendation result s12 is P 1,12 .

[0081] Table 2. First transition probability table of each intermediate event to the recommended event

[0082]

[0083]

[0084] Specifically, step 103 - 2 is implemented according to step 201 .

[0085] Step 201: Determine the products of the second transition probabilities and the target transition probabilities as third transition probabilities.

[0086] Step 103 - 3 is implemented according to step 202 .

[0087] Step 202: Determine the sum of the third transition probabilities as the probability of the intermediate event transitioning to the final recommendation results.

[0088] It can be understood that in the item recommendation chain, if the probability of intermediate event A transferring to intermediate event B is 0.3, and the probability of intermediate event B directly transferring to the final recommendation result Z is 0.6, then the probability of intermediate event A first transferring to intermediate event B and then directly transferring to the final recommendation result Z is 0.3*0.6=0.18. In other words, when intermediate event A occurs, the probability of intermediate event B and then finally reaching the final recommendation result Z is 0.18. Similarly, we can obtain the probability that intermediate event A first transfers to each intermediate event and then directly transfers to the final recommendation result Z. If there are three intermediate events A, B, and C in total, the probability that intermediate event A first transfers to intermediate event B and then directly transfers to the final recommendation result Z is 0.18, the probability that intermediate event A first transfers to intermediate event C and then directly transfers to the final recommendation result Z is 0.35, and the probability that intermediate event A directly transfers to the final recommendation result Z is 0.08. Then the probability that intermediate event A finally transfers to the final recommendation result Z is 0.18+0.35+0.08=0.56.

[0089] The following formula (1) is the calculation formula for the intermediate event A to be transferred to the intermediate event B first and then directly transferred from the intermediate event B to the intermediate event Z, where P A,B|总 is the second transition probability from intermediate event A to intermediate event B, P B,Z is the first transition probability of intermediate event B directly transferring to intermediate event Z; Formula (2) is the probability of intermediate event A finally transferring to the final recommendation result Z, where P A,Z|x is the second transition probability of intermediate event A to intermediate event x, and T is the total number of intermediate events.

[0090] P A,Z|B =P A,B|总 *P B,Z (1)

[0091]

[0092] Optionally, step 103 - 1 can be implemented according to the following steps 203 to 204 .

[0093] Step 203: Determine the second transition sub-probability of the intermediate event being transferred to the i-th intermediate event through N steps based on each of the first transition probabilities, where N ranges from 0 to M, and M is the number of transition steps corresponding to when the corresponding second transition sub-probability is not greater than a preset threshold.

[0094] Step 204: Determine the sum of the M+1 second transfer sub-probabilities as the second transfer probability of transferring the targeted intermediate event to the i-th intermediate event, where i ranges from 1 to T, and T is the total number of the intermediate events.

[0095] It is understandable that to calculate the second transition sub-probability of intermediate event A transitioning to intermediate event B, that is, the maximum probability of intermediate event A transitioning to intermediate event B, the corresponding second transition sub-probability for each number of transition steps from intermediate event A to intermediate event B can be calculated when the number of transition steps ranges from 0 to infinity. Intermediate event A and intermediate event B are any of the intermediate events, and intermediate event A and intermediate event B can be the same intermediate event or two different intermediate events.

[0096] In a specific embodiment, when the number of transition steps is 0, i.e., no transition occurs, the targeted intermediate event remains in its own intermediate event. In other words, the second transition sub-probability of the targeted intermediate event transitioning to the targeted intermediate event is 1, and the second transition sub-probability of the targeted intermediate event transitioning to an intermediate event other than the targeted intermediate event is 0. Specifically, when the number of transition steps is 0, the second transition sub-probability of intermediate event A transitioning to intermediate event A is 1, and the second transition sub-probability of intermediate event A transitioning to any intermediate event other than intermediate event A is 0.

[0097] When the number of transfer steps is infinite, since each transfer step requires multiplying the probability of transferring to the current intermediate event by the probability of transferring to the next intermediate event, for example, the transfer path from intermediate event A to intermediate event B after 5 steps is: A→C→A→B→C→B, then the second sub-transition probability of intermediate event A transferring to intermediate event B after 5 steps is: P A,C *P C,A *P A,B *P B,C *P C,B , P A,C The first transition probability represents the direct transition from intermediate event A to intermediate event C. Each first transition probability ranges from [0, 1]. Therefore, when the number of transition steps approaches infinity, the second transition probability of intermediate event A transitioning to intermediate event B after an infinite number of transition steps approaches 0. It can be considered that when the number of transition steps is infinite, the transition probability of intermediate event A transitioning to intermediate event B is 0. Therefore, M is the number of transition steps corresponding to the corresponding second transition probability not exceeding the preset threshold. In essence, it is the number of transition steps when the second transition probability approaches 0, that is, the preset threshold set by the server is infinitely close to 0.

[0098] In this way, when the number of transition steps ranges from 0 to M, the second transition sub-probability corresponding to each transition step of intermediate event A transferring to intermediate event B is obtained, then the second transition probability (that is, the maximum transition probability) that intermediate event A can transfer to intermediate event B is the sum of the second transition sub-probabilities corresponding to each transition step of intermediate event A transferring to intermediate event B.

[0099] The following formula (3) is the calculation formula for the second transition probability of intermediate event A to intermediate event B, P A,B|总 is the second transition probability of intermediate event A to intermediate event B, where P A,B|N is the probability that intermediate event A transfers to intermediate event B through N steps of transfer. M is infinitely close to infinity, so M can be represented by ∞.

[0100]

[0101] It can be understood that since prop recommendations are random, and the absorbing Markov chain is suitable for analyzing the potential laws of state transitions in random models, in the embodiment of the present application, the prop recommendation link can be mapped to the absorbing Markov chain to analyze the recommendation event transfer process in the prop recommendation scenario and the recommendation effect corresponding to the prop recommendation link.

[0102] A Markov chain is a discrete-time random process in mathematics that exhibits the Markov property. In this process, given current knowledge or information, the past (i.e., historical states before the current period) is irrelevant to predicting the future (i.e., future states after the current period). The Markov property is a concept in probability theory that states that, given the current state and all past states, the conditional probability distribution of a random process's future states depends only on the current state. In other words, given the current state, it is conditionally independent of the past states (i.e., the historical path of the process). This random process exhibits the Markov property, and processes exhibiting the Markov property are often called Markov processes.

[0103] The state transitions of a Markov chain are characterized by both zero aftereffect and ergodicity. Zero aftereffect means that the future state of an event depends solely on its current state, independent of its past state. Ergodicity means that starting from the current transition probability, after a number of transitions, the final probability of each state converges to a stable value. For state x, if, after reaching state x, the event remains permanently in x and cannot transition to another state, then state x is an absorbing state of the Markov chain; otherwise, it is a non-absorbing state. A Markov chain is called an absorbing Markov chain if it contains at least one absorbing state and an absorbing state can be reached from each non-absorbing state.

[0104] The transfer matrix corresponding to the prop recommendation link can be determined based on the probability of each intermediate event directly transferring to each recommended event in Table 2. As shown in the following matrix W, it is the transfer matrix of the prop recommendation link provided by the embodiment of the present application. The matrix W is a square matrix with T+L rows and columns, where T is the total number of intermediate events and L is the total number of final recommendation results. The element P in the matrix W is x2,y2 represents the one-step transition probability of directly transferring from the recommended event s-x2 to s-y2.

[0105]

[0106] In practical applications, the element P in the transfer matrix W x2,y2 The value of has the following rules: when s-x2 is the final recommendation result and s-y2 is the intermediate event, P x2,y2 =0; when s-x2 and s-y2 are the same recommended event and both are final recommendation results, P x2,y2 =1; when s-x2 and s-y2 are different recommended events and both are final recommendation results, P x2,y2 =0; when s-x2 and s-y2 are the same recommended event and both are intermediate events, P x2,y2 =0.

[0107] Among them, the matrix W can be split into four matrices Matrix Q is a T×T matrix, R is an L×T matrix, O is a T×L zero matrix, I is an L×L identity matrix, T is the total number of intermediate events, and L is the total number of final recommendation results. The item recommendation chain is mapped to an absorbing Markov chain and split into different matrices for subsequent analysis.

[0108] Therefore, in the absorbing Markov chain, step 203 can be implemented as follows:

[0109] Step 203 - 1 : Determine an intermediate event transition matrix according to each of the first transition probabilities.

[0110] Step 203 - 2 : Determine the element in the hth row and ith column of the Nth power matrix of the intermediate event transfer matrix as the second transfer sub-probability of the intermediate event transferring to the i-th intermediate event through N steps.

[0111] Accordingly, step 204 can be implemented according to the following steps:

[0112] Step 204 - 1 : Determine the sum of the elements in the h th row and the i th column in the M+1 N-th matrixes as the second transition probability of the intermediate event being targeted to transition to the i th intermediate event.

[0113] Among them, the elements in the kth row and jth column of the intermediate event transfer matrix represent the first transition probability of the kth intermediate event directly transferring to the jth intermediate event, wherein k and j are positive integers not greater than T, the hth row is the row where the intermediate event is located in the intermediate event transfer matrix, and the i-th column is the column where the i-th intermediate event is located in the intermediate event transfer matrix.

[0114] It can be understood that the intermediate event transfer matrix Q is a square matrix of T×T, and each element P in Q x3,y3 represents the first transition probability of directly transitioning from intermediate event s-x3 to intermediate event s-y3. In the example of the present application, the elements in matrix Q are the elements in the first nine rows and first nine columns of matrix W. As shown in the following matrix Q, it is an intermediate event transition matrix provided by an embodiment of the present application.

[0115]

[0116] To calculate the second transition probability from intermediate event s1 to intermediate event s5 for matrix Q, we can calculate the second transition sub-probabilities of intermediate event s1 to intermediate event s5 when the number of transition steps of intermediate event s1 increases from 0 to M. The sum of these second transition sub-probabilities is used as the second transition probability from intermediate event s1 to intermediate event s5.

[0117] Specifically, when the number of transfer steps is 0, the second transfer probability of the intermediate event s1 transferring to the intermediate event s5 is 0; when the number of transfer steps is 1, the second transfer probability of the intermediate event s1 transferring to the intermediate event s5 is P 1,5 ; When the number of transfer steps is 2, we can first obtain the probability of transferring from intermediate event s1 to each intermediate event when the number of transfer steps is 1, and then directly transfer from each intermediate event to intermediate event s5 when transferring to the second step. That is, when the number of transfer steps is 2, the transfer path can be: s1→s1→s5, s1→s2→s5, s1→s3→s5, s1→s4→s5, s1→s5→s5, s1→s6→s5, s1→s7→s5, s1→s8→s5, s1→s9→s5, then the second transfer sub-probability of intermediate event s1 transferring to intermediate event s5 is P 1,1 *P 1,5 +P 1,2 *P 2,5 +P 1,3 *P 3,5 +P 1,4 *P 4,5 +P 1,5 *P 5,5 +P 1,6 *P 6,5 +P 1,7 *P 7,5 +P1,8 *P 8,5 +P 1,9 *P 9,5 ; By analogy, we can obtain the second transition sub-probability when the number of transition steps is M, that is, when the intermediate event s1 transfers to the intermediate event s5.

[0118] As shown in Table 3, it is an example table of the number of transition steps, transition paths, and transition probabilities from the intermediate event s1 to the intermediate event s5 provided in the embodiment of the present application.

[0119] Table 3. Number of transition steps, transition paths, and transition probabilities from intermediate event s1 to intermediate event s5

[0120]

[0121] It can be seen that when the number of transfer steps is 0, the second transfer sub-probability corresponding to the transfer of intermediate event s1 to intermediate event s5 is the element 0 in the first row and fifth column of the zero-power matrix of matrix Q; when the number of transfer steps is 1, the second transfer sub-probability corresponding to the transfer of intermediate event s1 to intermediate event s5 is the element P in the first row and fifth column of matrix Q. 1,5 ; When the number of transfer steps is 2, the second transfer sub-probability corresponding to the intermediate event s1 transferring to the intermediate event s5 is the sum of the products of the elements in the 1st row and the elements in the 5th column in the matrix Q, that is, the elements in the 1st row and 5th column in the matrix obtained after matrix Q*matrix Q; similarly, when the number of transfer steps is M, the second transfer sub-probability corresponding to the intermediate event s1 transferring to the intermediate event s5 is the elements in the 1st row and 5th column in the M-th power matrix of matrix Q.

[0122] Therefore, the second transition probability of the intermediate event s1 to the intermediate event s5 is the sum of M+1 of the second transition sub-probabilities, which is essentially Q 0 +Q 1 +Q 2 …+Q M Then we get the element in row 1 and column 5 of the matrix V. The following formula (4) is the calculation formula of the matrix V. Each element P in the matrix V is x4,y4 represents the second transition probability from the intermediate event s-x4 to the intermediate event s-y4.

[0123]

[0124] Since the calculation process of the matrix V in Formula 4 is tedious and complicated, in order to simplify the calculation process and make the calculation of each second transition probability more efficient, the above Formula 4 can be converted into the following Formula (5) based on the calculation formula of the stable matrix of the absorbing Markov chain, where the matrix E is a unit matrix of the same order as the matrix Q.

[0125] V=(EQ)-1 (5)

[0126] Therefore, in a specific implementation, step 204-1 can be implemented as follows:

[0127] Step 205: Determine the inverse matrix of the matrix obtained by subtracting the intermediate event transfer matrix from the unit matrix as the first matrix V; determine the first matrix V as the sum matrix after summing M+1 N-th power matrices; determine the element in the h-th row and i-th column of the sum matrix as the second transition probability of the intermediate event being targeted to transfer to the i-th intermediate event.

[0128] By using the above method, the calculation process of the second transition probability of each intermediate event transferring to each intermediate event is simplified, the calculation pressure is reduced, and the calculation of the second transition probability is made more efficient.

[0129] Accordingly, in the absorbing Markov chain, step 201 can be implemented as follows:

[0130] A third matrix is ​​determined based on the target transition probabilities. The element in the i-th row and f-th column of the third matrix represents the first transition probability of the i-th intermediate event directly transferring to the f-th final recommendation result, where f is a positive integer not greater than L, and L is the total number of the final recommendation results.

[0131] The product of the elements in the hth row and ith column of the first matrix and the elements in the ith row and fth column of the third matrix is ​​determined as the third transition probability of transferring the targeted intermediate event to the i-th intermediate event and directly from the i-th intermediate event to the f-th final recommendation result.

[0132] Step 202 can be implemented by following the steps below:

[0133] determining a fourth matrix as a product of the first matrix and the third matrix;

[0134] The element in the hth row and the fth column of the fourth matrix is ​​determined as the probability of the intermediate event being transferred to the fth final recommendation result.

[0135] It can be understood that the third matrix R is a matrix of T×L, and each element P in R x5,y5 represents the first transition probability of transitioning directly from the intermediate event s-x5 to the final recommendation result s-y5. In the example of this application, the elements in matrix R are the elements in the first nine rows and last three columns of matrix W. The following matrix R is shown as the third matrix provided in the embodiment of this application.

[0136]

[0137] Combining matrix V and matrix R, if you want to calculate the probability that the intermediate event s1 will eventually transfer to the final recommended result s10, you can use the second transition probabilities of the intermediate event s1 transferring to the intermediate events s1~s9 multiplied by the first transition probability of each transferred intermediate event transferring to the final recommended result s10, and obtain the third transition probabilities of the intermediate event s1 first transferring to the intermediate events s1~s9 and then transferring to the final recommended result s10 from the transferred intermediate event.

[0138] In specific calculations, the probability that the intermediate event s1 is eventually transferred to the final recommendation result s10 is calculated as the sum of the products of the first row elements of the matrix V and the first column elements of the matrix R. Similarly, the probability that the intermediate event s1 is eventually transferred to the final recommendation result s11 is calculated as the sum of the products of the first row elements of the matrix V and the second column elements of the matrix R. Therefore, the probability that each intermediate event is eventually transferred to each final recommendation result is the element in the matrix obtained by multiplying the matrix V and the matrix R. That is, the first matrix V is multiplied by the third matrix R to obtain the fourth matrix. Each element P in the fourth matrix x6,y6 Represents the probability of transferring from the intermediate event s-x6 to the final recommendation result s-y6.

[0139] The embodiment of the present application maps the prop recommendation link to the state transition matrix of the corresponding absorbing Markov chain, and combines the derivation process of the absorbing Markov chain to conveniently, quickly and efficiently calculate the probability of each final recommendation result under each prop recommendation link.

[0140] Optionally, step 101 may be implemented as follows:

[0141] For each of the intermediate events, obtain the first number of occurrences that directly transfer to the qth recommended event from the total number of occurrences of the intermediate event within a preset historical time period, and determine the ratio of the first number of occurrences to the total number as the first transition probability of the intermediate event directly transferring to the qth recommended event, q traverses 1 to T+L.

[0142] It is understood that, for statistical convenience, in specific implementations, the total number of occurrences of each recommended event in the historical data within a historical time period can be directly counted. From the total number of occurrences of each recommended event, the first number of occurrences of a certain intermediate event directly transitioning to the qth recommended event within the historical time period can be obtained. The first transition probability of the intermediate event directly transitioning to the qth recommended event can be obtained based on the ratio of the first number of occurrences of the intermediate event directly transitioning to the qth recommended event to the total number of occurrences of the intermediate event.

[0143] For example, if 1,000 historical data are selected for item recommendation, the item s2 occurred 400 times, s2→s3 occurred 260 times, s2→s4 occurred 100 times, s2→s5 occurred 10 times, s2→s6 occurred 10 times, s2→s7 occurred 8 times, s2→s8 occurred 5 times, s2→s9 occurred 2 times, s2→s12 occurred 5 times, and s2→s1, s2→s2, s2→s9 and s2→s10 were all rejected 0 times.

[0144] Specifically, the normalized calculation is performed on the percentage of other recommended events that occur directly when each intermediate event occurs, so that the first transition probability of s2→s3 (directly selecting the associated alternative item when the target item is not found) is obtained as The probability of s2 directly transferring to other recommended events is similar to the calculation method of the first transition probability of s2 directly transferring to s3, which will not be repeated here.

[0145] By using the above method, after acquiring historical data, the first transition probability of each intermediate event directly transitioning to each recommended event can be quickly obtained, making the calculation of the first transition probability more efficient.

[0146] Optionally, the method for analyzing the recommendation effect of the item recommendation link provided in the embodiment of the present application may further include the following steps:

[0147] Step 301: In response to an adjustment instruction for an intermediate event to be adjusted in the item recommendation link, the intermediate event to be adjusted is adjusted corresponding to the adjustment instruction to obtain an adjusted item recommendation link.

[0148] Step 302: Based on the adjusted prop recommendation link, execute the step of obtaining the prop recommendation link in step 101 until the probability of each intermediate event transferring to each of the final recommendation results is obtained. The adjustment instruction includes: adding and / or deleting the intermediate event that is directly transferred to the intermediate event to be adjusted.

[0149] It is understood that after calculating and statistically determining the probabilities of each intermediate event leading to each final recommendation result, the developer can analyze the calculated statistical results. If the probability of user rejection and / or item non-launching corresponding to intermediate event A is high, the developer can adjust intermediate event A. Specifically, the probability of user rejection and item non-launching corresponding to each intermediate event can be screened out to find a probability that is not less than a preset probability value, and the intermediate event corresponding to the screened probability can be triggered to trigger an adjustment instruction.

[0150] The triggered adjustment instruction can be to add an intermediate event that can be directly transferred to the intermediate event to be adjusted, or to delete an intermediate event among the intermediate events to which the intermediate event to be adjusted can be directly transferred. For example, when the target prop is not found, a prop that is directly transferred to the recommended teammates frequently used can be added, and the props frequently used by teammates are marked as s13. The recommended minimum props in the intermediate event that can be directly transferred to when the target prop is not found can also be deleted.

[0151] It can be understood that when deleting the intermediate event to which the intermediate event to be adjusted can be directly transferred, you can first analyze the user rejections and / or props that have not been launched corresponding to each intermediate event. For example, when recommending a guaranteed prop, the probability of user rejection is high, then the recommended guaranteed prop can be deleted from the intermediate event to which the intermediate event to be adjusted can be directly transferred.

[0152] After receiving the adjustment instruction triggered by the developer for the intermediate event to be adjusted in the prop recommendation link, the intermediate event to be adjusted is adjusted corresponding to the adjustment instruction to obtain the adjusted prop recommendation link, and then step 101 and subsequent steps are executed until the probability of each intermediate event transferring to each final recommendation result is obtained.

[0153] It can be understood that in the absorbing Markov chain, the adjustment of the prop recommendation link is reflected in the adjustment of the rows and columns of the transfer matrix. Therefore, the analysis of the recommendation effect of the prop recommendation link before and after the adjustment in the absorbing Markov chain is simpler, more convenient, faster and more efficient.

[0154] Specifically, in the absorbing Markov chain, step 302 can be implemented by following the steps below:

[0155] According to the adjustment information of the adjusted prop recommendation link, corresponding rows and columns are added and / or deleted in the intermediate event transfer matrix to obtain the adjusted intermediate event transfer matrix; based on the adjusted intermediate event transfer matrix, the step of determining the inverse matrix of the matrix obtained by subtracting the intermediate event transfer matrix from the unit matrix in step 205 as the first matrix is ​​executed until the probability of each intermediate event transferring to each of the final recommendation results is obtained.

[0156] In the transfer matrix W of the absorbing Markov chain corresponding to the prop recommendation link, adding and / or deleting intermediate events corresponds to adding or deleting rows and columns in the transfer matrix. For example, if an intermediate event s13 is added: recommending props frequently used by teammates, the matrix W will have one row and one column added accordingly.

[0157] And based on the adjusted intermediate transfer matrix, step 205 and subsequent steps are executed until an adjusted fourth matrix representing the probability of each intermediate event transferring to each of the final recommendation results is obtained. Compare the values ​​of the first column (s10-user acceptance) of the fourth matrix before and after adjustment to see whether they are increased, whether the values ​​of the second column (s11-user rejection) are decreased, and whether the values ​​of the third column (s12-no props are launched) are decreased. This is used to better improve the prop recommendation link based on the analyzed results.

[0158] Through this technical means, the prop recommendation link is mapped to the absorbing Markov chain, and the transfer path and transfer probability of each intermediate event that ultimately reaches the final recommendation result are displayed in a method with clear logic and good visualization effect. In this way, when developers adjust the prop recommendation link, they can adjust the rows and columns of the transfer matrix W in the absorbing Markov chain, allowing developers to conveniently and efficiently analyze the recommendation effect of the recommendation link before and after the adjustment to obtain a more appropriate recommendation link.

[0159] like Figure 4 As shown, it is a transition matrix diagram determined by each first transition probability in the historical data provided by the embodiment of the present application. Matrix W is the transition matrix of the prop recommendation link. Rows 1 to 12 in matrix W represent the recommended events before the one-step transition occurs, and columns 1 to 12 in matrix W represent the recommended events transferred to after the one-step transition occurs. That is, each element in matrix W represents the probability of transferring from the recommended event corresponding to the row to the recommended event corresponding to the column. Matrix W can be split into four matrices: matrix Q, matrix R, matrix O, and matrix I. According to the stable matrix calculation formula of the absorbing Markov chain, the fourth matrix D can be obtained. Rows 1 to 9 in matrix D represent the intermediate events before the transition occurs, and columns 1 to 3 in matrix D represent the final recommended results transferred to after the transition occurs. That is, each element in matrix D represents the probability of transferring from the intermediate event corresponding to the row to the final recommended result corresponding to the column.

[0160] Corresponding to the analysis method of the prop recommendation link recommendation effect provided in the first embodiment of the present application, the second embodiment of the present application also provides an analysis device for the prop recommendation link recommendation effect, such as Figure 5 As shown, the device includes:

[0161] An acquisition unit 501 is configured to acquire an item recommendation link, wherein the item recommendation link includes recommendation events, including intermediate events and final recommendation results after the recommendation is completed. The intermediate events represent intermediate situations between the start and end of the recommendation.

[0162] A determining unit 502 is configured to determine, for each intermediate event, first transition probabilities of the intermediate event directly transitioning to each recommended event;

[0163] A calculation unit 503 is configured to calculate, for each intermediate event, a probability of the intermediate event transitioning to each of the final recommendation results based on each of the first transition probabilities;

[0164] The calculation unit 503 is specifically used to: determine the second transition probabilities of the intermediate event being targeted transferring to each of the intermediate events based on each of the first transition probabilities; determine the third transition probabilities based on each of the second transition probabilities and each of the target transition probabilities; the target transition probability is the first transition probability of each of the intermediate events directly transferring to each of the final recommendation results in each of the first transition probabilities, and the third transition probability is the probability that the intermediate event being targeted transfers to each of the intermediate events and then directly transfers from the transferred intermediate event to each of the final recommendation results; determine the probability that the intermediate event being targeted transfers to each of the final recommendation results based on each of the third transition probabilities.

[0165] Optionally, the calculation unit 503 is further specifically used to: determine the second transfer sub-probability of the intermediate event being transferred to the i-th intermediate event through N steps based on each of the first transfer probabilities, N traverses 0 to M, and M is the number of transfer steps corresponding to the corresponding second transfer sub-probability is not greater than a preset threshold; determine the sum of M+1 second transfer sub-probabilities as the second transfer probability of the intermediate event being transferred to the i-th intermediate event, i traverses 1 to T, and T is the total number of each of the intermediate events.

[0166] Optionally, the calculation unit 503 is further specifically used to: determine the intermediate event transfer matrix based on each first transfer probability, the elements in the kth row and jth column of the intermediate event transfer matrix represent the first transition probability of the kth intermediate event directly transferring to the jth intermediate event, wherein k and j are positive integers not greater than T; determine the elements in the hth row and i-th column of the Nth power matrix of the intermediate event transfer matrix as the second transfer sub-probability that the targeted intermediate event transfers to the i-th intermediate event through N steps, the hth row is the row where the targeted intermediate event is located in the intermediate event transfer matrix, and the i-th column is the column where the i-th intermediate event is located in the intermediate event transfer matrix.

[0167] The calculation unit 503 is further specifically configured to determine the sum of the elements in the h-th row and the i-th column in the M+1 N-th matrices as the second transition probability of the intermediate event being transferred to the i-th intermediate event.

[0168] Optionally, the calculation unit 503 is further specifically used to: determine the inverse matrix of the matrix obtained by subtracting the intermediate event transfer matrix from the unit matrix as the first matrix; determine the first matrix as the sum matrix after summing M+1 N-th power matrices; and determine the element in the h-th row and i-th column of the sum matrix as the second transition probability of the intermediate event being targeted to transfer to the i-th intermediate event.

[0169] Optionally, the calculation unit 503 is specifically used to: determine the product of each second transition probability and each target transition probability as each third transition probability; and determine the sum of each third transition probabilities as the probability of the intermediate event being transferred to each final recommendation result.

[0170] Optionally, the calculation unit 503 is also specifically used to: determine a third matrix based on each target transition probability, the elements in the i-th row and f-th column of the third matrix represent the first transition probability that the i-th intermediate event directly transfers to the f-th final recommendation result, wherein f is a positive integer not greater than L, and L is the total number of each of the said final recommendation results; and determine the product of the elements in the h-th row and i-th column of the first matrix and the elements in the i-th row and f-th column of the third matrix as the third transition probability that the targeted intermediate event transfers to the i-th intermediate event and then directly transfers from the i-th intermediate event to the f-th final recommendation result.

[0171] The calculation unit 503 is further configured to: determine the product of the first matrix and the third matrix as a fourth matrix; and determine the element in the hth row and the fth column of the fourth matrix as the probability of the intermediate event being transferred to the fth final recommendation result.

[0172] Optionally, the determination unit 502 is specifically used to: for each of the intermediate events, obtain the first number of occurrences that directly transfer to the qth recommended event from the total number of occurrences of the intermediate event within a preset historical time period, and determine the ratio of the first number of occurrences to the total number as the first transition probability of the intermediate event directly transferring to the qth recommended event, q traverses 1 to T+L.

[0173] Optionally, the device also includes an adjustment unit, which is used to: respond to an adjustment instruction for the intermediate event to be adjusted in the prop recommendation link, adjust the intermediate event to be adjusted corresponding to the adjustment instruction, and obtain an adjusted prop recommendation link; based on the adjusted prop recommendation link, execute the step of obtaining the prop recommendation link until the probability of each intermediate event transferring to each of the final recommendation results is obtained, and the adjustment instruction includes: adding and / or deleting the intermediate event directly transferred by the intermediate event to be adjusted.

[0174] Optionally, the adjustment unit is specifically used to: add and / or delete corresponding rows and columns in the intermediate event transfer matrix according to the adjustment information of the adjusted prop recommendation link to obtain the adjusted intermediate event transfer matrix; based on the adjusted intermediate event transfer matrix, perform the step of determining the inverse matrix of the matrix obtained by subtracting the intermediate event transfer matrix from the unit matrix as the first matrix, until the probability of each intermediate event transferring to each of the final recommendation results is obtained.

[0175] Corresponding to the analysis method of the prop recommendation link recommendation effect provided in the first embodiment of the present application, the third embodiment of the present application also provides an electronic device for running the analysis method of the prop recommendation link recommendation effect. Figure 6 As shown, the electronic device includes: a processor 601; and a memory 602, which is used to store a program for analyzing the recommendation effect of a prop recommendation link. After the device is powered on and the program for analyzing the recommendation effect of a prop recommendation link is run by the processor, the following steps are performed:

[0176] Obtaining a prop recommendation link, wherein the prop recommendation link includes various recommendation events, including intermediate events and final recommendation results after the recommendation is completed. The intermediate events are used to represent the intermediate situations between the start and end of the recommendation;

[0177] For each of the intermediate events, determining first transition probabilities of the intermediate event directly transitioning to each of the recommended events;

[0178] For each of the intermediate events, calculating, based on each of the first transition probabilities, a probability of the intermediate event transitioning to each of the final recommendation results;

[0179] The calculating, according to each first transition probability, the probability of the intermediate event being transferred to each final recommendation result includes:

[0180] Determining second transition probabilities of transferring the targeted intermediate event to the intermediate events according to the first transition probabilities;

[0181] Determine each third transition probability based on each second transition probability and each target transition probability; the target transition probability is each first transition probability of each intermediate event directly transitioning to each final recommendation result in each first transition probability, and the third transition probability is the probability that the targeted intermediate event transitions to each intermediate event and then directly transitions from the transferred intermediate event to each final recommendation result;

[0182] The probability of the intermediate event transitioning to each of the final recommendation results is determined according to each of the third transition probabilities.

[0183] Corresponding to the method for analyzing the recommendation effect of a prop recommendation link provided in the first embodiment of the present application, the fourth embodiment of the present application provides a computer-readable storage medium storing a program for analyzing the recommendation effect of a prop recommendation link, which is executed by a processor to perform the following steps:

[0184] Obtaining a prop recommendation link, wherein the prop recommendation link includes various recommendation events, including intermediate events and final recommendation results after the recommendation is completed. The intermediate events are used to represent the intermediate situations between the start and end of the recommendation;

[0185] For each of the intermediate events, determining first transition probabilities of the intermediate event directly transitioning to each of the recommended events;

[0186] For each of the intermediate events, calculating, based on each of the first transition probabilities, a probability of the intermediate event transitioning to each of the final recommendation results;

[0187] The calculating, according to each first transition probability, the probability of the intermediate event being transferred to each final recommendation result includes:

[0188] Determining second transition probabilities of transferring the targeted intermediate event to the intermediate events according to the first transition probabilities;

[0189] Determine each third transition probability based on each second transition probability and each target transition probability; the target transition probability is each first transition probability of each intermediate event directly transitioning to each final recommendation result in each first transition probability, and the third transition probability is the probability that the targeted intermediate event transitions to each intermediate event and then directly transitions from the transferred intermediate event to each final recommendation result;

[0190] The probability of the intermediate event transitioning to each of the final recommendation results is determined according to each of the third transition probabilities.

[0191] It should be noted that for the detailed description of the devices, electronic devices and computer-readable storage media provided in the second, third and fourth embodiments of the present application, reference can be made to the relevant description of the first embodiment of the present application, and no further details will be given here.

[0192] Although the present application is disclosed as above with the preferred embodiments, it is not intended to limit the present application. Any person skilled in the art may make possible changes and modifications without departing from the spirit and scope of the present application. Therefore, the scope of protection of the present application shall be based on the scope defined by the claims of the present application.

[0193] In a typical configuration, a node device in a blockchain includes one or more processors (CPUs), input / output interfaces, network interfaces, and memory.

[0194] Memory may include non-permanent storage in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. Memory is an example of a computer-readable medium.

[0195] 1. Computer-readable media includes permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), random access memory (RAM) of other types, read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage media, or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer-readable media does not include non-transitory media such as modulated data signals and carrier waves.

[0196] 2. Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware. Furthermore, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0197] Although the present application is disclosed as above with the preferred embodiments, it is not intended to limit the present application. Any person skilled in the art may make possible changes and modifications without departing from the spirit and scope of the present application. Therefore, the scope of protection of the present application shall be based on the scope defined by the claims of the present application.

Claims

1. A method for analyzing the recommendation effect of a prop recommendation link, characterized in that: The method comprises: Obtaining a prop recommendation link, wherein the prop recommendation link includes various recommendation events, including intermediate events and final recommendation results after the recommendation is completed. The intermediate events are used to represent the intermediate situations between the start and end of the recommendation; For each of the intermediate events, determining first transition probabilities of the intermediate event directly transitioning to each of the recommended events; For each of the intermediate events, calculating, based on each of the first transition probabilities, a probability of the intermediate event transitioning to each of the final recommendation results; The calculating, according to each first transition probability, the probability of the intermediate event being transferred to each final recommendation result includes: Determining second transition probabilities of transferring the targeted intermediate event to the intermediate events according to the first transition probabilities; Determine each third transition probability based on each second transition probability and each target transition probability; the target transition probability is each first transition probability of each intermediate event directly transitioning to each final recommendation result in each first transition probability, and the third transition probability is the probability that the targeted intermediate event transitions to each intermediate event and then directly transitions from the transferred intermediate event to each final recommendation result; The probability of the intermediate event transitioning to each of the final recommendation results is determined according to each of the third transition probabilities.

2. The method according to claim 1, characterized in that The determining, according to each of the first transition probabilities, each of the second transition probabilities of the intermediate event transitioning to each of the intermediate events includes: Determine, based on each of the first transition probabilities, a second transition sub-probability of the intermediate event being transferred to the i-th intermediate event through N steps, where N ranges from 0 to M, and M is the number of transition steps corresponding to when the corresponding second transition sub-probability is not greater than a preset threshold; The sum of the M+1 second transfer sub-probabilities is determined as the second transfer probability of the intermediate event being transferred to the i-th intermediate event, where i ranges from 1 to T, and T is the total number of the intermediate events.

3. The method according to claim 2, characterized in that The determining, according to each of the first transition probabilities, a second transition sub-probability that the intermediate event is transferred to the i-th intermediate event through N steps includes: Determine an intermediate event transfer matrix based on each of the first transfer probabilities, where the element in the kth row and jth column of the intermediate event transfer matrix represents a first transition probability of the kth intermediate event directly transferring to the jth intermediate event, where k and j are positive integers not greater than T; Determine the element in the h-th row and the i-th column of the N-th power matrix of the intermediate event transfer matrix as the second transfer sub-probability of the intermediate event being transferred to the i-th intermediate event through N steps, the h-th row being the row where the intermediate event is located in the intermediate event transfer matrix, and the i-th column being the column where the i-th intermediate event is located in the intermediate event transfer matrix; Determining the sum of the M+1 second transfer sub-probabilities as the second transfer probability of transferring the targeted intermediate event to the i-th intermediate event includes: The sum of the elements in the h-th row and the i-th column in the M+1 N-th matrixes is determined as the second transition probability of the intermediate event being transferred to the i-th intermediate event.

4. The method according to claim 3, characterized in that The step of determining the sum of the elements in the h-th row and the i-th column in the M+1 N-th matrixes as the second transition probability of the intermediate event being targeted to the i-th intermediate event includes: Determine the inverse matrix of a matrix obtained by subtracting the intermediate event transfer matrix from the identity matrix as the first matrix; Determine the first matrix as a sum matrix obtained by summing M+1 N-th power matrices; The element in the hth row and the ith column in the sum matrix is ​​determined as the second transition probability of the intermediate event being transferred to the i-th intermediate event.

5. The method according to claim 4, characterized in that The determining of each third transition probability according to each second transition probability and each target transition probability includes: Determine the product of each second transition probability and each target transition probability as each third transition probability; Determining the probability of the intermediate event being transferred to each of the final recommendation results according to each of the third transfer probabilities includes: The sum of the third transition probabilities is determined as the probability of the intermediate event transitioning to the final recommendation results.

6. The method according to claim 5, characterized in that Determining the products of the second transition probabilities and the target transition probabilities as third transition probabilities includes: Determine a third matrix based on each target transition probability, where the element in the i-th row and f-th column of the third matrix represents the first transition probability of the i-th intermediate event directly transitioning to the f-th final recommendation result, where f is a positive integer not greater than L, and L is the total number of the final recommendation results; The product of the element in the hth row and ith column of the first matrix and the element in the ith row and fth column of the third matrix is ​​determined as the third transition probability of the intermediate event being transferred to the i-th intermediate event and then directly transferred from the i-th intermediate event to the f-th final recommendation result; Determining the sum of the third transition probabilities as the probability of the intermediate event transitioning to the final recommendation results includes: determining a fourth matrix as a product of the first matrix and the third matrix; The element in the hth row and the fth column of the fourth matrix is ​​determined as the probability of the intermediate event being transferred to the fth final recommendation result.

7. The method according to claim 6, characterized in that The step of determining, for each intermediate event, first transition probabilities of the intermediate event directly transitioning to each recommended event includes: For each of the intermediate events, obtain the first number of occurrences that directly transfer to the qth recommended event from the total number of occurrences of the intermediate event within a preset historical time period, and determine the ratio of the first number of occurrences to the total number as the first transition probability of the intermediate event directly transferring to the qth recommended event, q traverses 1 to T+L.

8. The method according to claim 1, characterized in that Each of the intermediate events includes at least one of the following: Target item found, target item not found, related alternative items recommended, popular alternative items recommended, alternative items not existing, alternative items exceeding user quota cannot be launched, alternative items cannot be launched in the current scenario, and guaranteed items recommended; Each of the final recommendation results includes at least one of the following: user acceptance, user rejection, and no item being released.

9. The method according to claim 4, characterized in that The method further comprises: In response to an adjustment instruction for an intermediate event to be adjusted in the item recommendation link, adjusting the intermediate event to be adjusted corresponding to the adjustment instruction to obtain an adjusted item recommendation link; The step of obtaining the prop recommendation link is executed based on the adjusted prop recommendation link until the probability of each intermediate event transferring to each final recommendation result is obtained, and the adjustment instruction includes: adding and / or deleting the intermediate event directly transferred by the intermediate event to be adjusted.

10. The method according to claim 9, characterized in that The step of obtaining the item recommendation link based on the adjusted item recommendation link until the probability of each intermediate event transferring to each final recommendation result is obtained includes: According to the adjustment information of the adjusted item recommendation link, corresponding rows and columns are added and / or deleted in the intermediate event transfer matrix to obtain an adjusted intermediate event transfer matrix; Based on the adjusted intermediate event transfer matrix, the step of determining the inverse matrix of the matrix obtained by subtracting the intermediate event transfer matrix from the unit matrix as the first matrix is ​​performed until the probability of each intermediate event transferring to each final recommendation result is obtained.

11. A device for analyzing the recommendation effect of a prop recommendation link, characterized in that: The device comprises: An acquisition unit is configured to acquire an item recommendation link, wherein the item recommendation link includes recommendation events, wherein the recommendation events include intermediate events and final recommendation results after the recommendation is completed. The intermediate events are used to represent intermediate situations occurring between the start and end of the recommendation. a determining unit, configured to determine, for each of the intermediate events, first transition probabilities of the intermediate event directly transitioning to the recommended events; a calculation unit, configured to calculate, for each of the intermediate events, a probability of the intermediate event transitioning to each of the final recommendation results according to each of the first transition probabilities; The calculation unit is specifically used to: determine the second transition probabilities of the intermediate event being targeted transferring to each of the intermediate events based on the first transition probabilities; determine the third transition probabilities based on the second transition probabilities and the target transition probabilities; the target transition probabilities are the first transition probabilities of each of the intermediate events being directly transferred to each of the final recommendation results in the first transition probabilities, and the third transition probabilities are the probabilities of the intermediate event being targeted transferring to each of the intermediate events and then being directly transferred from the transferred intermediate event to each of the final recommendation results; determine the probability of the intermediate event being targeted transferring to each of the final recommendation results based on the third transition probabilities.

12. An electronic device, characterized in that: include: processor; as well as The memory is used to store a data processing program. After the electronic device is powered on and the program is run by the processor, the method according to any one of claims 1 to 10 is executed.

13. A computer-readable storage medium, characterized in that A data processing program is stored, and the program is run by a processor to perform the method according to any one of claims 1 to 10.

Images