CN111582895A - Product gender preference prediction method, system, device and storage medium - Google Patents

Abstract

The invention discloses a method, system, device and storage medium for predicting product gender preference. The method includes: acquiring user behavior data; the user behavior data includes: browsing the time series of products; inputting the user behavior data into an offline gender prediction model; the output parameters of the offline gender prediction model include: the gender value of the product; when judging the output of the offline gender prediction model When the gender value of the product is within the preset range, input the time series into the real-time gender prediction model; the output parameters of the real-time gender prediction model include: the gender value of the product; predict the gender attribute of the product according to the gender value output by the real-time gender prediction model. The invention utilizes a stable offline gender prediction model and a highly flexible real-time gender prediction model to predict the user's gender preference for a product, which is not only relatively stable, but can also accurately predict the user's current preference when the user changes preference.

Description

Prediction method, system, device and storage medium for product gender preference

technical field

The present invention relates to the field of Internet technology, and in particular, to a method, system, device and storage medium for predicting gender preference of products.

Background technique

In recent years, with the rapid development of e-commerce platforms, product personalized recommendation technology has also been greatly improved. Among them, the core content of the product recommendation system is the recommendation method, that is, how to accurately recommend products that are consistent with user interests to users.

At present, product recommendation methods mainly include content-based recommendation. In this recommendation method, the system obtains products that are consistent with users' purchasing interests from product feature information (product price, product type, product order quantity, etc.) through machine learning methods. Among them, the gender attribute of the product is one of the important features that affects the effect of the recommendation system. If women's clothing is recommended to male users, and men's bags to female users, it will seriously affect the user's shopping experience, so accurately predict the user's gender preference for products It is one of the key factors that determine the quality of the recommender system.

In the prior art, the historical behavior data of the user's business details page is used as a training sample, the training model is used to obtain a gender prediction model, the click behavior data of the target user's browsing product pages within a period of time is collected, and the gender prediction model is input to predict the target user's preference. Product gender preference. Since users' gender preferences for products are not constant, when users change their preferences, the gender prediction model cannot accurately predict users' current preferences. And for the gender prediction model, the model needs to be retrained at an appropriate time in order to fit the current application scene in a timely manner, which is very inconvenient to use.

SUMMARY OF THE INVENTION

The technical problem to be solved by the embodiments of the present invention is to provide a method for predicting the gender preference of a product in order to overcome the defect that the gender prediction model of the prior art is used to predict the gender preference of a user for a product with low accuracy and often needs to be retrained. , systems, devices and storage media.

The embodiment of the present invention solves the above-mentioned technical problems through the following technical solutions:

A method for predicting gender preference of a product, the method for predicting gender preference includes:

Obtaining user behavior data; the user behavior data includes: time series of browsing products; the products are provided with signs representing gender attributes;

Inputting the user behavior data into an offline gender prediction model; the output parameters of the offline gender prediction model include: the gender value of the product; the gender value represents the gender attribute of the product;

Judging whether the gender value output by the offline gender prediction model is within a preset range;

When it is judged to be yes, input the time series into a real-time gender prediction model; the output parameters of the real-time gender prediction model include: the gender value of the product;

The gender attribute of the product is predicted according to the gender value output by the real-time gender prediction model.

Preferably, when the judgment is no, the gender attribute of the product is predicted according to the gender value output by the offline gender prediction model.

Preferably, the output parameters of the real-time gender prediction model further include: preference;

The calculation formula of the preference is as follows:

Among them, RV represents preference; 1≤i≤k; k represents the number of products in the time series; genderValue _i represents the gender attribute of the ith product; t _i represents the decay function; α represents the decay parameter.

Preferably, the step of acquiring user behavior data specifically includes:

Obtain the user behavior data based on Kafka;

And/or, after the step of predicting the gender attribute of the product, further comprising:

Store the prediction results in Redis.

An electronic device, comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the prediction of product gender preference as described in any of the above when the processor executes the computer program method.

A computer-readable storage medium having a computer program stored thereon, when the computer program is executed by a processor, implements the steps of the method for predicting gender preference of a product according to any one of the above.

A prediction system for product gender preference, the prediction system includes:

a data acquisition module for acquiring user behavior data; the user behavior data includes: a time series of browsing products; the products are provided with signs representing gender attributes;

a calculation module, used to input the user behavior data into an offline gender prediction model; the output parameters of the offline gender prediction model include: the gender value of the product; the gender value represents the gender attribute of the product;

a judgment module for judging whether the gender value output by the offline gender prediction model is within a preset range, and when the judgment is yes, calling the calculation module;

The computing module is further configured to input the time series into a real-time gender prediction model; the output parameters of the real-time gender prediction model include: the gender value of the product;

The prediction module is used to predict the gender attribute of the product according to the gender value output by the real-time gender prediction model.

Preferably, when the judgment module judges no, the prediction module is further configured to predict the gender attribute of the product according to the gender value output by the offline gender prediction model.

Preferably, the output parameters of the real-time gender prediction model further include: preference;

The calculation formula of the preference is as follows:

Among them, RV represents preference; 1≤i≤k; k represents the number of products in the time series; genderValue _i represents the gender attribute of the ith product; t _i represents the decay function; α represents the decay parameter.

Preferably, the data acquisition module is specifically configured to acquire the user behavior data based on Kafka;

And/or, the prediction system further includes: a storage module;

The storage module is used for storing the prediction result in Redis.

The positive improvement effect of the embodiment of the present invention is that: the embodiment of the present invention uses a stable offline gender prediction model and a highly flexible real-time gender prediction model to predict the user's gender preference for a product, which is not only more stable, but also when the user changes the preference tendency. , it can also accurately predict the user's current preferences, and without retraining the model, it can fit the current application scene and is easy to use.

Description of drawings

FIG. 1 is a flowchart of a method for predicting product gender preference according to Embodiment 1 of the present invention.

FIG. 2 is a schematic structural diagram of an electronic product according to Embodiment 2 of the present invention.

FIG. 3 is a schematic block diagram of a product gender preference prediction system according to Embodiment 4 of the present invention.

Detailed ways

The present invention is further described below by way of examples, but the present invention is not limited to the scope of the described examples.

Example 1

This embodiment provides a method for predicting a user's gender preference for a product. As shown in Figure 1, the predicting method includes the following steps:

Step 101: Obtain user behavior data.

The user behavior data includes: user ID, SKU (stock keeping unit), time series of user browsing (clicking) products, and the like. The time series of products browsed by users includes an array of gender attributes of k (may but not limited to 50) products whose browsing time is closest to the current moment; for each product, there is an identifier, which is used to characterize the gender attribute of the product, for example , respectively use 0, 0.5, 1 to represent men's products, neutral products (both men and women can use) and women's products.

In this embodiment, streaming data collection is specifically used to extract fields required to predict user preferences, such as user ID, SKU, and time series of products browsed by users, from kafka (an open source stream processing platform) log.

Step 102: Input the user behavior data into the offline gender prediction model.

Among them, the output parameters of the offline gender prediction model include the gender value and preference of the product. The gender value represents the gender attribute of the product that the user may purchase at the next moment. The preference degree represents the possibility that the user purchases products with gender attributes predicted by the offline gender prediction model. For example, the gender value of the product output by the offline gender prediction model is 0.3 (representing male products), and the preference degree is 0.88, indicating that the user’s gender value is predicted to be 0.88. The probability of preference for men's products at one time is 0.88.

The following is a brief description of the training process of the offline gender prediction model:

Make a basic table of browsing (click) behavior of users' business details page in units of days, hours or weeks, and extract the time series of user history (such as the previous day) browsing products as training data and test data. Specifically, using the sliding window method, the attribute data of the last browsed product is used as the test data, and the attribute data of the previously browsed product is used as the training data by sliding windows of different sizes (3 times, 5 times, 10 times of behavior data) . Calculate the gender ratio in each sequence for feature extraction. If there are 5 males, 3 females, and 2 neutrals in ten times, the feature extraction is male 5/10, female 3/10, and neutral 2/10. According to the feature extraction results, a multiclass logistic regression training model is performed to obtain an offline gender prediction model.

Step 103: Determine whether the gender value output by the offline gender prediction model falls within a preset range.

If it is judged to be yes, it means that the gender preference of the user predicted by the offline gender prediction model is not obvious, the prediction result is not ideal, and further calculation is required, then go to step 104, and replace the preference output by the real-time gender prediction model with the offline gender prediction model. The preference of the gender prediction model output. If it is determined to be no, it means that the prediction result is ideal, and no further calculation is required, and step 105 is executed.

In this embodiment, the preset range is obtained according to multiple random simulation experiments, and different preset ranges can be set for different usage scenarios. For example, the preset range in this embodiment adopts [0.4, 0.6].

Step 104: Input the time series of browsing products into the real-time gender prediction model.

The output parameters of the real-time gender prediction model include: product gender value, preference, confidence, and the like.

The formula for calculating preference is as follows:

Among them, RV represents preference; range(1, 50) represents the number of gender attribute arrays; 1≤i≤50; genderValue _i represents the gender attribute of the i-th product, and 0, 0.5, and 1 are used to represent male products, neutral Articles (both male and female) and female articles; t _i represents the decay function; α represents the decay parameter, α∈(0.5,1), the closer α is to 1, the greater the decay amplitude. Because the closer the browsing time of the product is, the stronger the gender possibility of users preferring this product, and the weaker the longer the time. Therefore, the decay function will give a higher weight to the product at the most recent moment in the browsing time, and the weight of the product in the future will gradually decrease.

Step 105: Predict the gender attribute of the product according to the gender value and output the prediction result.

The prediction result includes: gender attribute, preference degree, confidence degree, etc., and the preference degree is the preference degree output by the real-time gender prediction model.

The formula for calculating the gender attribute of a product is as follows:

Among them, f(gender) represents the gender attribute function; p _male (confidence of male products) represents the proportion of products whose gender attribute is male in the time series of browsing products; gender _male represents the gender attribute of male; p _female (female product) p middle (confidence of neutral products) indicates that in the time series of browsing products, the gender attribute of products is female; gender _female indicates that the gender attribute is female; p _middle (confidence of neutral products) indicates that in the time series of browsing products, the gender attribute is The proportion of products that are neutral; gender _middle indicates that the gender attribute is neutral.

It should be noted that it is assumed that the user clicked and browsed 10 products, including 4 male products, 4 female products, and 2 neutral products; then p _male = 0.4, p _female = 0.4, p _middle = 0.2, Since the number of clicks on male and female products by the user is the same, indicating that the user has no special gender preference, the gender attribute of the user is predicted to be neutral.

Step 106: Store the prediction result in Redis.

In this embodiment, the prediction model is arranged in a Storm cluster (a server cluster with a master-slave structure). When making predictions, Kafka sends the target user's behavior data to the Storm cluster for data processing and model prediction, and finally stores the prediction results in Redis (a storage system).

In this embodiment, using a stable offline gender prediction model and a highly flexible real-time gender prediction model to predict the user's gender preference for a product is not only relatively stable, but also accurately predicts the user's current preference when the user changes preference. , and without retraining the model, it can fit the current application scene and is easy to use.

Example 2

FIG. 2 is a schematic structural diagram of an electronic device according to an embodiment of the present invention, and shows a block diagram of an exemplary electronic device 90 suitable for implementing the embodiments of the present invention. The electronic device 90 shown in FIG. 2 is only an example, and should not impose any limitation on the function and scope of use of the embodiments of the present invention.

As shown in FIG. 2, the electronic device 90 may take the form of a general-purpose computing device, which may be, for example, a server device. The components of the electronic device 90 may include, but are not limited to: the above-mentioned at least one processor 91 , the above-mentioned at least one memory 92 , and a bus 93 connecting different system components (including the memory 92 and the processor 91 ).

The bus 93 includes a data bus, an address bus and a control bus.

Memory 92 may include volatile memory, such as random access memory (RAM) 921 and/or cache memory 922 , and may further include read only memory (ROM) 923 .

The memory 92 may also include a program tool 925 (or utility tool) having a set (at least one) of program modules 924 including, but not limited to, an operating system, one or more application programs, other program modules, and programs Data, each or some combination of these examples may include an implementation of a network environment.

The processor 91 executes various functional applications and data processing by running the computer program stored in the memory 92, such as the method for predicting the gender preference of a product provided in Embodiment 1 of the present invention.

The electronic device 90 may also communicate with one or more external devices 94 (eg, keyboards, pointing devices, etc.). Such communication may take place through input/output (I/O) interface 95 . Also, the model-generated electronic device 90 may also communicate with one or more networks (eg, a local area network (LAN), a wide area network (WAN), and/or a public network such as the Internet) through a network adapter 96 . As shown, network adapter 96 communicates with other modules of model-generated electronics 90 via bus 93 . It should be understood that, although not shown in the figures, other hardware and/or software modules may be used in conjunction with the model-generated electronics 90, including but not limited to: microcode, device drivers, redundant processors, external disk drive arrays, RAID ( disk arrays) systems, tape drives, and data backup storage systems.

It should be noted that although several units/modules or sub-units/modules of the electronic device are mentioned in the above detailed description, this division is merely exemplary and not mandatory. Indeed, the features and functions of two or more units/modules described above may be embodied in one unit/module according to embodiments of the present invention. Conversely, the features and functions of one unit/module described above may be further subdivided to be embodied by multiple units/modules.

Example 3

This embodiment provides a computer-readable storage medium on which a computer program is stored, and when the program is executed by a processor, implements the steps of the method for predicting gender preference of a product provided in Embodiment 1.

Wherein, the readable storage medium may include, but is not limited to, a portable disk, a hard disk, a random access memory, a read-only memory, an erasable programmable read-only memory, an optical storage device, a magnetic storage device, or any of the above suitable combination.

In a possible implementation manner, the present invention can also be implemented in the form of a program product, which includes program codes, when the program product runs on a terminal device, the program code is used to cause the terminal device to execute the implementation The steps of the method for predicting product gender preference described in Example 1.

Wherein, the program code for executing the present invention can be written in any combination of one or more programming languages, and the program code can be completely executed on the user equipment, partially executed on the user equipment, as an independent The software package executes on the user's device, partly on the user's device, partly on the remote device, or entirely on the remote device.

Example 4

This embodiment provides a prediction system for a user's gender preference for a product. As shown in FIG. 3 , the prediction system includes: a data acquisition module 1 , a calculation module 2 , a judgment module 3 , a prediction module 4 , and a storage module 5 .

The data acquisition module 1 is used for acquiring user behavior data. The user behavior data includes: user ID, SKU (stock keeping unit), time series of user browsing (clicking) products, and the like. The time series of products browsed by users includes an array of gender attributes of k (may but not limited to 50) products whose browsing time is closest to the current moment; for each product, there is an identifier, which is used to characterize the gender attribute of the product, for example , respectively use 0, 0.5, 1 to represent men's products, neutral products (both men and women can use) and women's products.

In this embodiment, the data acquisition module 1 is specifically used to acquire user behavior data based on Kafka, that is, to use streaming data collection to extract fields needed to predict user preferences from Kafka logs, such as user ID, SKU, and user browsing time for products sequence etc.

The calculation module 2 is used for inputting the user behavior data into the offline gender prediction model. Among them, the output parameters of the offline gender prediction model include the gender value and preference of the product. The gender value represents the gender attribute of the product that the user may purchase at the next moment. The preference degree represents the possibility that the user purchases products with gender attributes predicted by the offline gender prediction model. For example, the gender value of the product output by the offline gender prediction model is 0.3 (representing male products), and the preference degree is 0.88, indicating that the user’s gender value is predicted to be 0.88. The probability of preference for men's products at one time is 0.88. The training process of the offline gender prediction model is similar to the model training process in the prediction method shown in Embodiment 1, and will not be repeated here.

The judgment module 3 is used to judge whether the gender value output by the offline gender prediction model is within the preset range; if the judgment is no, it means that the prediction result is ideal, and the prediction module 4 is called to predict the product according to the gender value output by the offline gender prediction model. If it is judged to be yes, it means that the gender preference of the user predicted by the offline gender prediction model is not obvious, the prediction result is not ideal, and further calculation is needed, then the calculation module 2 is called.

In this embodiment, the preset range is obtained according to multiple random simulation experiments, and different preset ranges can be set for different usage scenarios. For example, the preset range in this embodiment adopts [0.4, 0.6].

The calculation module 2 is also used to input the time series of the browsed products into the real-time gender prediction model, and call the prediction module 4 to predict the gender attribute of the product according to the gender value output by the real-time gender prediction model and output the prediction result. The output parameters of the real-time gender prediction model include: product gender value, preference, confidence, etc., and the preference is the preference output by the real-time gender prediction model.

The formula for calculating preference is as follows:

Among them, RV represents preference; range(1,50) represents the number of gender attribute arrays in the time series; 1≤i≤50; genderValue _i represents the gender attribute of the i-th product, and 0, 0.5, and 1 are used to represent male products. , neutral products (both male and female) and female products; t _i represents the decay function; α represents the decay parameter, α∈(0.5, 1), the closer α is to 1, the greater the decay amplitude. Because the closer the browsing time of the product is, the stronger the gender possibility of users preferring this product, and the weaker the longer the time. Therefore, the decay function will give a higher weight to the product at the most recent moment in the browsing time, and the weight of the product in the future will gradually decrease.

In this embodiment, the prediction result output by the prediction module 4 includes: gender attribute, preference degree, confidence degree, and the like.

The formula for calculating the gender attribute of a product is as follows:

Among them, f(gender) represents the gender attribute function; p _male (confidence of male products) represents the proportion of products whose gender attribute is male in the time series of browsing products; gender _male represents the gender attribute of male; p _female (female product) p middle (confidence of neutral products) indicates that in the time series of browsing products, the gender attribute of products is female; gender _female indicates that the gender attribute is female; p _middle (confidence of neutral products) indicates that in the time series of browsing products, the gender attribute is The proportion of products that are neutral; gender _middle indicates that the gender attribute is neutral.

The storage module 5 is used to store the prediction result in Redis.

In this embodiment, the prediction model is arranged in a Storm cluster (a server cluster with a master-slave structure). When making predictions, Kafka sends the target user's behavior data to the Storm cluster for data processing and model prediction, and finally stores the prediction results in Redis (a storage system).

In this embodiment, using a stable offline gender prediction model and a highly flexible real-time gender prediction model to predict the user's gender preference for a product is not only relatively stable, but also accurately predicts the user's current preference when the user changes preference. , and without retraining the model, it can fit the current application scene and is easy to use.

Although the specific embodiments of the present invention are described above, those skilled in the art should understand that this is only an illustration, and the protection scope of the present invention is defined by the appended claims. Those skilled in the art can make various changes or modifications to these embodiments without departing from the principle and essence of the present invention, but these changes and modifications all fall within the protection scope of the present invention.

Claims (10)

1. A method for predicting product gender preference, which is characterized in that the method for predicting gender preference comprises the following steps:

acquiring user behavior data; the user behavior data includes: browsing a time series of products; the product is provided with an identifier representing gender attribute;

inputting the user behavior data into an offline gender prediction model; the output parameters of the off-line gender prediction model comprise: a gender number of the product; the gender value characterizes a gender attribute of the product;

judging whether the gender numerical value output by the off-line gender prediction model is in a preset range or not;

if yes, inputting the time sequence into a real-time gender prediction model; the output parameters of the real-time gender prediction model comprise: a gender number of the product;

and predicting the gender attribute of the product according to the gender numerical value output by the real-time gender prediction model.

2. The method of claim 1, wherein if the determination is negative, the gender attribute of the product is predicted according to the gender value outputted from the offline gender prediction model.

3. The method of predicting product gender preferences of claim 1 wherein the output parameters of said real-time gender prediction model further comprise: a preference degree;

the calculation formula of the preference degree is as follows:

wherein RV represents a preference; i is more than or equal to 1 and less than or equal to k; k represents the number of products in the time series; GenderValue_iCharacterizing a gender attribute of the ith product; t is t_iRepresenting the decay function, α representing the decay parameters.

4. The method for predicting the product property preferences of any one of claims 1-3, wherein the step of obtaining the user behavior data specifically comprises:

acquiring the user behavior data based on Kafka;

and/or, after the step of predicting the gender attribute of the product, further comprising:

the prediction result is stored in Redis.

5. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements the method of predicting productivity preferences of any of claims 1-4 when executing the computer program.

6. A computer-readable storage medium, on which a computer program is stored which, when being executed by a processor, carries out the steps of the method of predicting product preference according to any one of claims 1 to 4.

7. A system for predicting productivity preferences, the system comprising:

the data acquisition module is used for acquiring user behavior data; the user behavior data includes: browsing a time series of products; the product is provided with an identifier representing gender attribute;

a calculation module for inputting the user behavior data into an offline gender prediction model; the output parameters of the off-line gender prediction model comprise: a gender number of the product; the gender value characterizes a gender attribute of the product;

the judgment module is used for judging whether the gender numerical value output by the off-line gender prediction model is in a preset range or not, and calling the calculation module when the gender numerical value output by the off-line gender prediction model is judged to be in the preset range;

the computing module is further configured to input the time series into a real-time gender prediction model; the output parameters of the real-time gender prediction model comprise: a gender number of the product;

and the prediction module is used for predicting the gender attribute of the product according to the gender numerical value output by the real-time gender prediction model.

8. The system of claim 7, wherein the prediction module is further configured to predict gender attributes of the product according to the gender values outputted from the offline gender prediction model when the determination module determines no.

9. The system for predicting product gender preferences of claim 7 wherein the output parameters of said real-time gender prediction model further comprises: a preference degree;

the calculation formula of the preference degree is as follows:

wherein RV represents a preference; i is more than or equal to 1 and less than or equal to k; k represents the number of products in the time series; GenderValue_iCharacterizing a gender attribute of the ith product; t is t_iRepresenting the decay function, α representing the decay parameters.

10. The system for predicting product suitability preferences according to any one of claims 7-9 wherein the data acquisition module is specifically configured to acquire the user behavior data based on Kafka;

and/or, the prediction system further comprises: a storage module;

the storage module is used for storing the prediction result in Redis.

Images