Method, device and computer readable storage medium for determining a package

Abstract

Embodiments of the present application provide a method, device and computer readable storage medium for determining a package, the method comprising: obtaining a first target feature of N user groups in a historical period, N being a positive integer; for any one user group, obtaining a second target feature of the user group in a future period based on a target transfer probability algorithm and the first target feature of the user group in the historical period; and determining a corresponding charge package of each user group according to the second target feature of each user group in the future period and at least one machine learning model. Embodiments of the present application can improve the accuracy of determining the required charge package of a user group and improve user experience.

Description

Technical Field

[0001] This application belongs to the field of artificial intelligence technology, and in particular relates to a method, apparatus, device and computer-readable storage medium for determining a package. Background Technology

[0002] As the internet becomes increasingly integrated into society, the economy, and daily life, people have become accustomed to various online information interactions, generating massive amounts of data. This data contains a wealth of characteristic information, such as users' consumption behavior and basic information. These characteristics include users' consumption intentions and also reveal their potential spending power.

[0003] However, when recommending corresponding phone plans to users based on their feature data, the time lag in the user's feature data can affect the accuracy of the recommendation, resulting in a large deviation between the recommended phone plan and the user's actual needs. Summary of the Invention

[0004] This application provides a method, apparatus, device, and computer-readable storage medium for determining a call package, which can accurately determine the call package required by each user group and improve user experience.

[0005] In a first aspect, embodiments of this application provide a method for determining a service package. The method includes: obtaining first target features of N user groups in a historical period, where N is a positive integer; for any user group, obtaining second target features of the user group in a future period based on a target transition probability algorithm and the first target features of the user group in the historical period; and determining the corresponding call package for each user group based on the second target features of each user group in the future period and at least one machine learning model.

[0006] According to an implementation of the first aspect of this application, the first target feature includes x first discrete features and y first continuous features, and the second target feature includes x second discrete features and y second continuous features, where x and y are both positive integers; for any user group, based on the target transition probability algorithm and the first target features of the user group in the historical period, the second target features of the user group in the future period are obtained, including: based on the target transition probability algorithm and the x first discrete features of the user group in the historical period, x second discrete features of the user group in the future period are obtained; based on the target transition probability algorithm and the y first continuous features of the user group in the historical period, y second continuous features of the user group in the future period are obtained.

[0007] According to any of the foregoing embodiments of the first aspect of this application, the target transition probability algorithm includes a Markov chain transition probability algorithm, the historical period includes at least two sub-historical periods, and the user group includes a first user who handled the business corresponding to the first discrete feature in the sub-historical period and a second user who did not handle the business corresponding to the first discrete feature in the sub-historical period; based on the target transition probability algorithm and x first discrete features of the user group in the historical period, x second discrete features of the user group in future periods are obtained, including: for any first discrete feature, calculating a first probability that the first user in the i-th sub-historical period becomes the first user in the (i+1)-th sub-historical period, calculating a second probability that the first user in the i-th sub-historical period becomes the second user in the (i+1)-th sub-historical period, calculating a third probability that the second user in the i-th sub-historical period becomes the first user in the (i+1)-th sub-historical period, and calculating a fourth probability that the second user in the i-th sub-historical period becomes the second user in the (i+1)-th sub-historical period, where i is a positive integer; obtaining a first transition probability matrix corresponding to the first discrete feature based on the first probability, second probability, third probability, and fourth probability; and determining the second discrete feature corresponding to the first discrete feature based on the first transition probability matrix corresponding to the first discrete feature.

[0008] According to any of the foregoing embodiments of the first aspect of this application, before obtaining the first target features of N user groups in a historical period, the method further includes: classifying multiple users into multiple user groups based on target classification theory and target classification model, wherein the target classification theory includes survival, interrelationship and growth theory; for each type of user group, classifying each type of user group into at least one user group based on a clustering algorithm; and summarizing the user groups in the multiple user groups to obtain N user groups.

[0009] According to any of the foregoing embodiments of the first aspect of this application, before classifying multiple users into multiple user groups based on target classification theory and target classification model, the method further includes: obtaining the features of multiple users in a historical period; processing the features of multiple users in a historical period based on feature engineering to obtain the features of multiple users after feature engineering processing; obtaining the first target features of N user groups in a historical period, including: obtaining the first target features of N user groups in a historical period based on the features of multiple users after feature engineering processing.

[0010] According to any of the foregoing embodiments of the first aspect of this application, before processing the features of multiple users in a historical period based on feature engineering to obtain the features of multiple users after feature engineering processing, the method further includes: performing data cleaning on the features of multiple users in a historical period to obtain the features of multiple users after data cleaning; performing normalization processing on the features of multiple users after data cleaning to obtain the features of multiple users after normalization processing; processing the features of multiple users in a historical period based on feature engineering to obtain the features of multiple users after feature engineering processing includes: processing the features of multiple users after normalization processing based on feature engineering to obtain the features of multiple users after feature engineering processing.

[0011] According to any of the foregoing embodiments of the first aspect of this application, determining the corresponding call charge package for each user group based on the second target feature of each user group in the future period and at least one machine learning model includes: for any user group's second target feature in the future period, inputting the second target feature of the user group in the future period into multiple different machine learning models to obtain the prediction results of multiple machine learning models for the call charge package corresponding to the user group; when the prediction results of a preset number of machine learning models are consistent, the prediction results of the preset number of machine learning models are used as the call charge package corresponding to the user group.

[0012] According to any of the foregoing embodiments of the first aspect of this application, a variety of different machine learning models include at least two of the following: random forest model, backpropagation neural network model, support vector machine model, logistic regression model, distributed gradient boosting model, and K-nearest neighbor classification algorithm model.

[0013] Secondly, embodiments of this application provide a package determination device, which includes: a first acquisition module, used to acquire first target features of N user groups in a historical period, where N is a positive integer; a prediction module, used to obtain second target features of any user group in a future period based on a target transition probability algorithm and the first target features of the user group in the historical period; and a first determination module, used to determine the corresponding call package for each user group based on the second target features of each user group in the future period and at least one machine learning model.

[0014] Thirdly, embodiments of this application provide an electronic device, which includes: a processor, a memory, and a computer program stored in the memory and executable on the processor. When the computer program is executed by the processor, it implements the steps of the method for determining the package as provided in the first aspect.

[0015] Fourthly, embodiments of this application provide a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the method for determining the package as provided in the first aspect.

[0016] The method, apparatus, device, and computer-readable storage medium for determining service packages according to embodiments of this application first obtain the first target features of N user groups in a historical period, where N is a positive integer. Next, for any user group, based on a target transition probability algorithm and the first target features of the user group in the historical period, a second target feature of the user group in a future period is obtained. Then, based on the second target features of each user group in the future period and at least one machine learning model, the corresponding service package for each user group is determined. This embodiment of the application determines the corresponding service package for each user group based on the second target features of each user group in the future period predicted by the target transition probability algorithm and at least one machine learning model. This method utilizes the second target features of the user group in the future period to determine the corresponding service package, reducing the impact of the lag in user feature data on the determined service package, improving the accuracy of determining the required service package for each user group, and enhancing the user experience. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments of this application will be briefly introduced below. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 A flowchart illustrating a method for determining a package as provided in an embodiment of this application;

[0019] Figure 2 A schematic flowchart of step S102 in the method for determining the package provided in the embodiments of this application;

[0020] Figure 3 Another flowchart illustrating the method for determining the package provided in this application embodiment;

[0021] Figure 4 Another flowchart illustrating the method for determining the package provided in this application embodiment;

[0022] Figure 5 Another flowchart illustrating the method for determining the package provided in this application embodiment;

[0023] Figure 6 A schematic flowchart of step S103 in the method for determining the package provided in the embodiments of this application;

[0024] Figure 7 A schematic flowchart of a package determination device provided in an embodiment of this application;

[0025] Figure 8 A schematic diagram of the hardware structure of the electronic device provided in an embodiment of this application is shown. Detailed Implementation

[0026] The features and exemplary embodiments of various aspects of this application will be described in detail below. To make the objectives, technical solutions, and advantages of this application clearer, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only intended to explain this application and not to limit it. For those skilled in the art, this application can be implemented without some of these specific details. The following description of the embodiments is merely to provide a better understanding of this application by illustrating examples.

[0027] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising..." does not exclude the presence of additional identical elements in the process, method, article, or apparatus that includes said element.

[0028] It should be understood that the term "and / or" used in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Additionally, the character " / " in this article generally indicates that the preceding and following related objects have an "or" relationship.

[0029] Various modifications and variations can be made to this application without departing from its spirit or scope, which will be apparent to those skilled in the art. Therefore, this application is intended to cover modifications and variations falling within the scope of the corresponding claims (the claimed technical solutions) and their equivalents. It should be noted that the embodiments provided in this application can be combined with each other without contradiction.

[0030] Before describing the technical solutions provided in the embodiments of this application, in order to facilitate understanding of the embodiments of this application, this application first specifically explains the problems existing in the related technologies:

[0031] As the internet becomes increasingly integrated into society, the economy, and daily life, people have become accustomed to various online information interactions, generating massive amounts of data. This data contains a wealth of characteristic information, such as users' consumption behavior and basic information. These characteristics include users' consumption intentions and also reveal their potential spending power.

[0032] However, when recommending corresponding phone plans to users based on their feature data, the time lag in the user's feature data can affect the accuracy of the recommendation, resulting in a large deviation between the recommended phone plan and the user's actual needs.

[0033] To address the problems in the prior art, embodiments of this application provide a method, apparatus, device, and computer-readable storage medium for determining a package.

[0034] The technical concept of this application embodiment is as follows: First, obtain the first target features of N user groups in a historical period, where N is a positive integer; next, for any user group, based on the target transition probability algorithm and the first target features of the user group in the historical period, obtain the second target features of the user group in the future period; then, based on the second target features of each user group in the future period and at least one machine learning model, determine the corresponding call plan for each user group. This application embodiment determines the corresponding call plan for each user group based on the second target features of each user group in the future period predicted by the target transition probability algorithm and at least one machine learning model. This utilizes the second target features of the user group in the future period to determine the corresponding call plan, reducing the impact of the lag in user feature data on the determined call plan, improving the accuracy of determining the required call plan for the user group, and enhancing the user experience.

[0035] The method for determining the package provided in the embodiments of this application will be introduced first below.

[0036] Figure 1 This is a flowchart illustrating a method for determining a package as provided in an embodiment of this application. Figure 1 As shown, the method for determining the package may include the following steps S101 to S103.

[0037] S101. Obtain the first target features of N user groups in the historical period, where N is a positive integer.

[0038] A user group can include at least one user. The historical period refers to a past timeframe, such as the past month, the past three months, or the past six months. The primary target features include, but are not limited to, the user's age, gender, location, and features related to the user's spending history.

[0039] S102. For any user group, based on the target transition probability algorithm and the first target feature of the user group in the historical period, obtain the second target feature of the user group in the future period.

[0040] The target transition probability algorithm includes, but is not limited to, the Markov chain transition probability algorithm, which will be described in detail below and will not be repeated here. For any user group, the first target feature of the user group in the historical period can be predicted based on the target transition probability algorithm, and the changing trend of the first target feature in the future period can be predicted, thereby obtaining the second target feature of the user group in the future period. The future period is a future time period, such as the next month, the next three months, or the next six months.

[0041] S103. Based on the second target features of each user group in the future period and at least one machine learning model, determine the corresponding call charge package for each user group.

[0042] Among them, the call plans include, but are not limited to, 4G plans or 5G plans.

[0043] A machine learning model can be pre-built based on at least one machine learning algorithm and trained to obtain a trained machine learning model.

[0044] In S103, the second target features of each user group in the future period can be input into at least one machine learning model to obtain the result of the call charge package corresponding to each user group output by at least one machine learning model.

[0045] The method for determining the service package in this application embodiment first obtains the first target features of N user groups in a historical period, where N is a positive integer. Next, for any user group, based on a target transition probability algorithm and the first target features of the user group in the historical period, a second target feature of the user group in the future period is obtained. Then, based on the second target features of each user group in the future period and at least one machine learning model, the corresponding service package for each user group is determined. This application embodiment determines the corresponding service package for each user group based on the second target features of each user group in the future period predicted by the target transition probability algorithm and at least one machine learning model. This method utilizes the second target features of the user group in the future period to determine the corresponding service package, reducing the impact of the lag in user feature data on the determined service package, improving the accuracy of determining the required service package for each user group, and enhancing the user experience.

[0046] According to some embodiments of this application, optionally, the first target feature may include x first discrete features and y first continuous features. The second target feature may include x second discrete features and y second continuous features, where x and y are both positive integers.

[0047] Discrete features are features that can only take a finite number of values, such as gender or whether a certain transaction has been completed. Discrete features are usually represented by integers or strings and cannot take arbitrary values. Continuous features are features that can take any number of values, such as height or call duration. Continuous features can be represented by real numbers.

[0048] Figure 2 This is a schematic flowchart of step S102 in the method for determining the package provided in this application embodiment. For example... Figure 2 As shown, according to some embodiments of this application, optionally, S102, for any user group, obtaining the second target feature of the user group in the future period based on the target transition probability algorithm and the first target feature of the user group in the historical period, may include the following steps S201 and S202.

[0049] S201. Based on the target transition probability algorithm and x first discrete features of the user group in the historical period, x second discrete features of the user group in the future period are obtained.

[0050] Taking the Markov chain transition probability algorithm as an example, a Markov chain is a stochastic process in probability theory and mathematical statistics that possesses Markov properties and exists within a discrete exponent set and state space. It is represented by a conditional distribution, and P(Xn+1|Xn) is called the "transition probability" in the stochastic process. Through the model transformation of the Markov chain, the state of an event can be converted into a transition probability matrix (also known as a state distribution matrix). The historical period can include at least two sub-historical periods. The time span of each sub-historical period can be flexibly adjusted according to the actual situation, and this application does not limit this. For example, in some examples, the time span of a sub-historical period can be one month.

[0051] The user group can include a first user who subscribed to the service corresponding to the first discrete feature in the sub-historical period and a second user who did not subscribe to the service corresponding to the first discrete feature in the sub-historical period. For example, if the first discrete feature is "whether the user subscribed to a data package", the first user is the user who subscribed to a data package in the sub-historical period, and the second user is the user who did not subscribe to a data package in the sub-historical period.

[0052] In some specific embodiments, S201, based on the target transition probability algorithm and x first discrete features of the user group in the historical period, obtains x second discrete features of the user group in the future period, which may specifically include step one, step two and step three.

[0053] Step 1: For any first discrete feature, calculate the first probability that the first user in the i-th sub-historical period becomes the first user in the (i+1)-th sub-historical period, calculate the second probability that the first user in the i-th sub-historical period becomes the second user in the (i+1)-th sub-historical period, calculate the third probability that the second user in the i-th sub-historical period becomes the first user in the (i+1)-th sub-historical period, and calculate the fourth probability that the second user in the i-th sub-historical period becomes the second user in the (i+1)-th sub-historical period, where i is a positive integer.

[0054] For example, the first transformation probability matrix corresponding to the first discrete feature is:

[0055]

[0056] Among them, P 1,1 P represents the first probability that the first user in the i-th sub-historical period becomes the first user in the (i+1)-th sub-historical period. 1,2 P represents the second probability that the first user in the i-th sub-historical period will become the second user in the (i+1)-th sub-historical period. 2,1 P represents the third probability that the second user in the i-th sub-historical period becomes the first user in the (i+1)-th sub-historical period. 2,2This represents the fourth probability that the second user in the i-th sub-historical period will become the second user in the (i+1)-th sub-historical period.

[0057] Taking the first discrete feature as "whether the user subscribes to a data package" as an example, assuming 100,000 user samples are selected from the user group, if the user subscribes to data package X in the i-th sub-historical period... v The number of the first users is m v1 Then in the current period For m v1 / 100000. Next, calculate the subscription status of traffic packages in the (i+1)th sub-historical period. If traffic package X was subscribed in the (i+1)th sub-historical period... v The number of the first users is m v2 Then in the current period For m v2 / 100000. In some examples, the first probability P 1,1 It can be and The mean of the second probability P. 1,2 =1-P 1,1 .

[0058] If no data package X was ordered in the i-th sub-historical period v The number of second users is m v3 Then in the current period For m v3 / 100000. Next, calculate the subscription status of the traffic package in the (i+1)th sub-historical period. If no traffic package X was subscribed in the (i+1)th sub-historical period... v The number of the second users is m v4 Then in the current period For m v4 / 100000. In some examples, the fourth probability P 2,2 It can be and The mean. The third probability P. 2,1 =1-P 2,2 .

[0059] To make the obtained first transformation probability matrix more objective and accurate, in some embodiments, the first probability P can be determined based on the results of multiple sub-historical periods. 1,1 Second probability P 1,2 The third probability P 2,1 and the fourth probability P 2,2 .

[0060] For example, using the same calculation and Similarly, calculate the current period of the (i+2)th sub-historical period. Calculate the current period of the (i+3)th sub-historical period ... calculate the current period of the (i+n)th sub-historical period. n is an integer greater than 1. The first probability P 1,1 This can be within multiple sub-historical periods The corresponding mean, u = 1, 2, ..., n+1.

[0061] For example, using the same calculation and Similarly, calculate the current period of the (i+2)th sub-historical period. Calculate the current period of the (i+3)th sub-historical period ... calculate the current period of the (i+n)th sub-historical period. n is an integer greater than 1. The fourth probability P 2,2 This can be within multiple sub-historical periods The corresponding mean, u = 1, 2, ..., n+1.

[0062] The second probability P 1,2 =1-P 1,1 The third probability P 2,1 =1-P 2,2 .

[0063] Step 2: Based on the first probability, second probability, third probability and fourth probability, obtain the first transformation probability matrix corresponding to the first discrete feature.

[0064] After obtaining the first probability P 1,1 Second probability P 1,2 The third probability P 2,1 and the fourth probability P 2,2 Then, the first transformation probability matrix P corresponding to the first discrete feature can be obtained. (u,v) .

[0065] Step 3: Determine the second discrete feature corresponding to the first discrete feature based on the first transformation probability matrix corresponding to the first discrete feature.

[0066] Using the first discrete feature corresponding to the first transformation probability matrix P (u,v) This allows us to predict the second discrete feature of the next future cycle, such as predicting the order of traffic package X in the next future cycle. v The number of first users and the number of unsubscribed traffic packages X in the next future cycle v The number of second-generation users.

[0067] S202. Based on the target transition probability algorithm and the y first continuous features of the user group in the historical period, obtain the y second continuous features of the user group in the future period.

[0068] In S202, the Markov chain transition probability algorithm can be used to predict the first continuous features of the user group in the historical period y, and obtain the second continuous features of the user group in the future period y. The specific process is similar to that in S201, and will not be repeated here.

[0069] Further research by the inventors of this application revealed that current related technologies typically employ a single-layer segmentation strategy when classifying user groups. However, if the target sample size for a single clustering is large, the convergence speed is slow, and the clustering results are prone to converge to local optima rather than obtaining the global optimal solution. This makes the clustering effect susceptible to noise, edge points, and isolated points, and it is also poor for high-dimensional data objects.

[0070] In view of this, some embodiments of this application employ a two-layer cross-segmentation strategy, combining ERG theory (existence, relatedness, growth, i.e., the three core needs theory of survival, interdependence, and growth), the Kano model, and actual business needs to divide users into multiple user groups. Then, a clustering algorithm is used to precisely subdivide each user group into atomic user clusters (minimum user groups), achieving accurate user segmentation. The Kano model is a useful tool for classifying and prioritizing user needs.

[0071] Figure 3 This is another flowchart illustrating the method for determining the package provided in this application embodiment. For example... Figure 3 As shown, according to some embodiments of this application, optionally, before obtaining the first target characteristics of N user groups in the historical period in S101, the method for determining the package may further include the following steps S301 to S303.

[0072] S301. Based on target classification theory and target classification model, multiple users are divided into multiple user groups. Target classification theory includes survival, interrelationship and growth theory.

[0073] Target classification models include, but are not limited to, the Kano model. For example, in some embodiments, based on ERG theory and the Kano model, user needs can be categorized into five types from low to high: basic, urgent, stable, expected, and attractive. Correspondingly, multiple users can be divided into five user groups. These five user groups can be labeled with need type tags L1, L2, L3, L4, and L5, respectively.

[0074] S302. For each type of user group, the user groups are divided into at least one user group based on a clustering algorithm.

[0075] Clustering algorithms include, but are not limited to, the K-means clustering algorithm. For any user group, it can be further subdivided into at least one user group based on the clustering algorithm.

[0076] When dividing various user groups into at least one user group based on clustering algorithms, the user group division can be based on parameters such as user scale, user revenue, and communication consumption level, or by combining multiple parameters.

[0077] S303. Summarize the user groups from the multiple user groups to obtain N user groups.

[0078] Thus, this application adopts a two-layer cross-segmentation strategy, combining ERG theory, Kano model and actual business needs to divide users into multiple user groups, and then uses clustering algorithm to accurately subdivide atomic user groups (minimum user groups) in each user group, thereby achieving accurate user segmentation and helping to match accurate call plans for each segmented atomic user group.

[0079] Further research by the inventors of this application revealed that current related technologies primarily employ the random forest algorithm for filtering user features. However, the random forest algorithm can only filter features that significantly impact the prediction results. Features that significantly impact the prediction results do not necessarily mean that the features are effective. Incorporating ineffective features into the model expands the dimensions of data analysis, increasing data sparsity and modeling complexity.

[0080] In view of this, some embodiments of this application employ feature engineering methods such as analysis of variance, correlation tests, feature information value assessment, and chi-square tests to screen features. On the other hand, when the user samples for each package are imbalanced, the recommendation accuracy for rare categories is low. Therefore, embodiments of this application recommend packages to each atomic user group separately, thereby improving the recommendation accuracy for each category.

[0081] Figure 4 This is another flowchart illustrating the method for determining the package provided in this application embodiment. For example... Figure 4 As shown, according to some embodiments of this application, optionally, before S301, which divides multiple users into multiple user groups based on target classification theory and target classification model, the method for determining the package may further include the following steps S401 and S402.

[0082] S401. Obtain the characteristics of multiple users in a historical period.

[0083] That is, before classifying multiple users, you can first obtain the characteristics of multiple users in historical periods.

[0084] S402. Based on feature engineering, the features of multiple users in the historical period are processed to obtain the features of multiple users after feature engineering.

[0085] In some examples, feature engineering methods such as analysis of variance, correlation tests, information value (IV) assessment, and / or chi-square tests can be used to process the features of multiple users over a historical period, resulting in feature-engineered features for multiple users. This essentially involves feature filtering.

[0086] Accordingly, S201, obtaining the first target features of N user groups in a historical period includes the following steps:

[0087] Based on the features processed by multiple user feature engineering, the first target features of N user groups in the historical period are obtained.

[0088] Figure 5 This is another flowchart illustrating the method for determining the package provided in this application embodiment. For example... Figure 5 As shown, according to some embodiments of this application, optionally, before S402, processing the features of multiple users in a historical period based on feature engineering to obtain the features of multiple users after feature engineering processing, the method for determining the package may further include the following steps S501 and S502.

[0089] S501. Perform data cleaning on the characteristics of multiple users in historical periods to obtain the cleaned characteristics of multiple user data.

[0090] In S501, the latest user information records can be obtained, feature indicators can be extracted as necessary, invalid data can be filtered, and basic data processing such as missing values ​​and extreme values ​​can be completed.

[0091] S502. Normalize the features of multiple user data after cleaning to obtain the normalized features of multiple users.

[0092] In S502, data standardization can be performed. For example, Min-Max normalization can be used to limit the data to 0 to 1, ensuring that the feature information is dimensionless.

[0093] Accordingly, S402, processing the features of multiple users over a historical period based on feature engineering to obtain the features of multiple users after feature engineering processing, may include the following steps:

[0094] Based on feature engineering, the normalized features of multiple users are processed to obtain the features of multiple users after feature engineering.

[0095] According to some embodiments of this application, optionally, after S402, processing the features of multiple users in historical periods based on feature engineering to obtain the features of multiple users after feature engineering processing, the method for determining the package may further include the following steps:

[0096] Based on the scenario and / or business needs, determine m features that need to be used.

[0097] Where m is a positive integer. That is, in feature engineering, features can be further filtered according to the scenario and / or business needs to select the features that are finally needed.

[0098] Further research by the inventors of this application revealed that current related technologies typically employ a single machine learning model for prediction, which results in low accuracy of the prediction results.

[0099] In view of this, some embodiments of this application select and integrate various machine learning models with high prediction accuracy, such as random forest model, backpropagation neural network model, support vector machine model, logistic regression model, distributed gradient boosting model and K-nearest neighbor classification algorithm model. When the prediction results of a preset number of machine learning models are consistent, the final prediction result is output, thereby improving the accuracy of the prediction result.

[0100] Figure 6 This is a schematic flowchart of step S103 in the method for determining the package provided in this application embodiment. For example... Figure 6 As shown, according to some embodiments of this application, optionally, S103, determining the corresponding call charge package for each user group based on the second target features of each user group in the future period and at least one machine learning model, may include the following steps S601 and S602.

[0101] S601. For any user group's second target feature in the future period, input the user group's second target feature in the future period into multiple different machine learning models to obtain the prediction results of multiple machine learning models for the user group's corresponding call charge package.

[0102] The various machine learning models include at least two of the following: random forest, backpropagation neural network, support vector machine, logistic regression, distributed gradient boosting, and K-nearest neighbor classification algorithm. For example, in some embodiments, one of the six best-performing machine learning models can be used for prediction. The second target feature of the user group in the future period is input into these various machine learning models to obtain prediction results for the corresponding phone bill packages for the user group.

[0103] S602. When the prediction results of a preset number of machine learning models are consistent, the prediction results of the preset number of machine learning models shall be used as the corresponding call charge packages for the user group.

[0104] Machine learning models can be evaluated from four perspectives: accuracy, recall, precision, and F1 score. At least two machine learning models with high evaluation scores are selected. If the predictions of these two models are consistent, then the predictions of these two models are used as the corresponding phone plan for the user group.

[0105] Thus, in some embodiments of this application, multiple different machine learning models with high prediction accuracy, such as random forest model, backpropagation neural network model, support vector machine model, logistic regression model, distributed gradient boosting model and K-nearest neighbor classification algorithm model, are integrated. When the prediction results of a preset number of machine learning models are consistent, the final prediction result is output, thereby improving the accuracy of the prediction result.

[0106] Based on the method for determining the package provided in the above embodiments, this application also provides a specific implementation of a package determination device. Please refer to the following embodiments.

[0107] Figure 7 This is a schematic flowchart of a package determination device provided in an embodiment of this application. Figure 7 As shown, the package determination device 70 provided in this application embodiment may include the following modules:

[0108] The first acquisition module 701 is used to acquire the first target features of N user groups in a historical period, where N is a positive integer;

[0109] Prediction module 702 is used to obtain the second target feature of a user group in a future period based on the target transition probability algorithm and the first target feature of the user group in the historical period for any given user group.

[0110] The first determining module 703 is used to determine the corresponding call charge package for each user group based on the second target features of each user group in the future period and at least one machine learning model.

[0111] The package determination device of this application embodiment acquires first target features of N user groups in a historical period, where N is a positive integer; for any user group, based on a target transition probability algorithm and the first target features of the user group in the historical period, a second target feature of the user group in a future period is obtained; based on the second target features of each user group in the future period and at least one machine learning model, the corresponding call package for each user group is determined. This application embodiment determines the corresponding call package for each user group based on the second target features of each user group in the future period predicted by the target transition probability algorithm and at least one machine learning model. This method utilizes the second target features of the user group in the future period to determine the corresponding call package, reducing the impact of the lag in user feature data on the determined call package, improving the accuracy of determining the required call package for the user group, and enhancing the user experience.

[0112] In some embodiments, the first target feature includes x first discrete features and y first continuous features, and the second target feature includes x second discrete features and y second continuous features, where x and y are both positive integers. The prediction module 702 is specifically used to obtain x second discrete features of the user group in future periods based on the target transition probability algorithm and x first discrete features of the user group in historical periods; and to obtain y second continuous features of the user group in future periods based on the target transition probability algorithm and y first continuous features of the user group in historical periods.

[0113] In some embodiments, the target transition probability algorithm includes a Markov chain transition probability algorithm, the historical period includes at least two sub-historical periods, and the user group includes a first user who processed the business corresponding to the first discrete feature in the sub-historical period and a second user who did not process the business corresponding to the first discrete feature in the sub-historical period. The prediction module 702 is specifically used to, for any first discrete feature, calculate, for the first user in the i-th sub-historical period, a first probability that the first user in the i-th sub-historical period will become the first user in the (i+1)-th sub-historical period; calculate, for the first user in the i-th sub-historical period, a second probability that the first user in the i-th sub-historical period will become the second user in the (i+1)-th sub-historical period; calculate, for the second user in the i-th sub-historical period, a third probability that the second user in the i-th sub-historical period will become the first user in the (i+1)-th sub-historical period; and calculate, for the second user in the i-th sub-historical period, a fourth probability that the second user in the i-th sub-historical period will become the second user in the (i+1)-th sub-historical period, where i is a positive integer; obtain the first transition probability matrix corresponding to the first discrete feature based on the first probability, second probability, third probability, and fourth probability; and determine the second discrete feature corresponding to the first discrete feature based on the first transition probability matrix corresponding to the first discrete feature.

[0114] In some embodiments, the package determination device 70 provided in this application may further include a classification module, used to classify multiple users into multiple user groups based on target classification theory and target classification model, the target classification theory including survival, interrelationship and growth theory; for each type of user group, the user groups are divided into at least one user group based on a clustering algorithm; the user groups in the multiple user groups are summarized to obtain N user groups.

[0115] In some embodiments, the package determination device 70 provided in this application may further include a feature engineering processing module, used to acquire features of multiple users in a historical period; and to process the features of multiple users in the historical period based on feature engineering to obtain features of multiple users after feature engineering processing. The first acquisition module 701 is specifically used to obtain first target features of N user groups in a historical period based on the features of multiple users after feature engineering processing.

[0116] In some embodiments, the package determination device 70 provided in this application may further include a preprocessing module, used to clean the features of multiple users in a historical period to obtain cleaned features of multiple users; and to normalize the cleaned features of multiple users to obtain normalized features of multiple users. The feature engineering processing module is specifically used to process the normalized features of multiple users based on feature engineering to obtain feature-engineered features of multiple users.

[0117] In some embodiments, the first determining module 703 is specifically used to input the second target feature of any user group in the future period into multiple different machine learning models to obtain the prediction results of multiple machine learning models for the corresponding call charge package of the user group; when the prediction results of a preset number of machine learning models are consistent, the prediction results of the preset number of machine learning models are used as the corresponding call charge package of the user group.

[0118] In some embodiments, a variety of different machine learning models include at least two of the following: random forest model, backpropagation neural network model, support vector machine model, logistic regression model, distributed gradient boosting model, and K-nearest neighbor classification algorithm model.

[0119] Figure 7 Each module / unit in the device shown has the function of implementing each step in the method for determining the package provided in the above method embodiment, and can achieve its corresponding technical effect. For the sake of brevity, it will not be described in detail here.

[0120] Based on the method for determining the package provided in the above embodiments, this application also provides specific implementation methods for electronic devices. Please refer to the following embodiments.

[0121] Figure 8 A schematic diagram of the hardware structure of the electronic device provided in an embodiment of this application is shown.

[0122] Electronic devices may include a processor 801 and a memory 802 storing computer program instructions.

[0123] Specifically, the processor 801 may include a central processing unit (CPU), an application specific integrated circuit (ASIC), or one or more integrated circuits that can be configured to implement the embodiments of this application.

[0124] Memory 802 may include mass storage for data or instructions. For example, and not limitingly, memory 802 may include a hard disk drive (HDD), floppy disk drive, flash memory, optical disk, magneto-optical disk, magnetic tape, or Universal Serial Bus (USB) drive, or a combination of two or more of these. In one example, memory 802 may include removable or non-removable (or fixed) media, or memory 802 may be a non-volatile solid-state memory. Memory 802 may be internal or external to an electronic device.

[0125] In one example, memory 802 may be read-only memory (ROM). In one example, the ROM may be a mask-programmed ROM, a programmable ROM (PROM), an erasable PROM (EPROM), an electrically erasable PROM (EEPROM), an electrically rewritable ROM (EAROM), or flash memory, or a combination of two or more of these.

[0126] Memory 802 may include read-only memory (ROM), random access memory (RAM), disk storage media device, optical storage media device, flash memory device, electrical, optical, or other physical / tangible memory storage device. Therefore, typically, memory includes one or more tangible (non-transitory) computer-readable storage media (e.g., memory devices) encoded with software including computer-executable instructions, and when the software is executed (e.g., by one or more processors), it is operable to perform the operations described with reference to the method according to one aspect of this application.

[0127] The processor 801 reads and executes the computer program instructions stored in the memory 802 to implement the methods / steps in the above method embodiments and achieve the corresponding technical effects achieved by the method embodiments in executing their methods / steps. For the sake of brevity, these details will not be repeated here.

[0128] In one example, the electronic device may also include a communication interface 803 and a bus 810. For example, Figure 8 As shown, the processor 801, memory 802, and communication interface 803 are connected through bus 810 and complete communication with each other.

[0129] The communication interface 803 is mainly used to realize communication between various modules, devices, units and / or equipment in the embodiments of this application.

[0130] Bus 810 includes hardware, software, or both, that couples components of an electronic device together. For example, and not limitingly, the bus may include an Accelerated Graphics Port (AGP) or other graphics bus, an Extended Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), a Hyper Transport (HT) interconnect, an Industry Standard Architecture (ISA) bus, an Infinite Bandwidth Interconnect, a Low Pin Count (LPC) bus, a memory bus, a Microchannel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCI-X) bus, a Serial Advanced Technology Attachment (SATA) bus, a Video Electronics Standards Association Local (VLB) bus, or other suitable buses, or combinations of two or more of these. Where appropriate, bus 810 may include one or more buses. Although specific buses are described and illustrated in embodiments of this application, this application contemplates any suitable bus or interconnect.

[0131] Furthermore, in conjunction with the package determination method in the above embodiments, this application embodiment can provide a computer-readable storage medium for implementation. This computer-readable storage medium stores computer program instructions; when these computer program instructions are executed by a processor, they implement any of the package determination methods in the above embodiments. Examples of computer-readable storage media include non-transitory computer-readable storage media, such as electronic circuits, semiconductor memory devices, ROM, random access memory, flash memory, erasable ROM (EROM), floppy disks, CD-ROMs, optical disks, and hard disks.

[0132] It should be clarified that this application is not limited to the specific configurations and processes described above and shown in the figures. For the sake of brevity, detailed descriptions of known methods are omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of this application is not limited to the specific steps described and shown. Those skilled in the art can make various changes, modifications, and additions, or change the order of steps, after understanding the spirit of this application.

[0133] The functional blocks shown in the above-described block diagram can be implemented as hardware, software, firmware, or a combination thereof. When implemented in hardware, they can be, for example, electronic circuits, application-specific integrated circuits (ASICs), appropriate firmware, plug-ins, function cards, etc. When implemented in software, the elements of this application are programs or code segments used to perform the required tasks. Programs or code segments can be stored on a machine-readable medium or transmitted over a transmission medium or communication link via data signals carried on a carrier wave. "Machine-readable medium" can include any medium capable of storing or transmitting information. Examples of machine-readable media include electronic circuits, semiconductor memory devices, ROM, flash memory, erasable ROM (EROM), floppy disks, CD-ROMs, optical disks, hard disks, fiber optic media, radio frequency (RF) links, etc. Code segments can be downloaded via computer networks such as the Internet, intranets, etc.

[0134] It should also be noted that the exemplary embodiments mentioned in this application describe methods or systems based on a series of steps or apparatus. However, this application is not limited to the order of the above steps; that is, the steps can be performed in the order mentioned in the embodiments, or in a different order, or several steps can be performed simultaneously.

[0135] The aspects of this application have been described above with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It should be understood that each block in the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing apparatus to produce a machine such that these instructions, executable via the processor of the computer or other programmable data processing apparatus, enable the implementation of the functions / actions specified in one or more blocks of the flowchart illustrations and / or block diagrams. Such a processor can be, but is not limited to, a general-purpose processor, a special-purpose processor, a special application processor, or a field-programmable logic circuit. It is also understood that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can also be implemented by dedicated hardware performing the specified functions or actions, or can be implemented by a combination of dedicated hardware and computer instructions.

[0136] The above description is merely a specific implementation of this application. Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, modules, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here. It should be understood that the protection scope of this application is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this application, and these modifications or substitutions should all be covered within the protection scope of this application.

Claims

1. A method for determining a package deal, characterized in that, include: Based on target classification theory and target classification model, multiple users are divided into multiple user groups according to their needs. The target classification theory includes survival, interrelationship and growth theories. For any user group, based on at least one of the following factors—user size, user revenue, and communication consumption level—the user group will be further subdivided into at least one user group using a clustering algorithm. The user groups in the multiple user groups are aggregated to obtain N user groups; Obtain the first target features of N user groups in a historical period. The first target features include x first discrete features and y first continuous features, where N is a positive integer and x and y are both positive integers. For any user group, based on the target transition probability algorithm and the x first discrete features of the user group in the historical period, x second discrete features of the user group in the future period are obtained. Based on the target transition probability algorithm and the y first continuous features of the user group in the historical period, the y second continuous features of the user group in the future period are obtained. Based on the second target features of each user group in the future period and at least one machine learning model, the corresponding call charge package for each user group is determined. The second target features include the x second discrete features and the y second continuous features.

2. The method according to claim 1, characterized in that, The target transition probability algorithm includes the Markov chain transition probability algorithm, the historical period includes at least two sub-historical periods, and the user group includes a first user who handles the business corresponding to the first discrete feature in the sub-historical period and a second user who does not handle the business corresponding to the first discrete feature in the sub-historical period. The process of obtaining x second discrete features of the user group in future periods based on the target transition probability algorithm and the x first discrete features of the user group in historical periods includes: For any of the first discrete features, calculate the first probability that the first user in the i-th sub-historical period becomes the first user in the (i+1)-th sub-historical period, calculate the second probability that the first user in the i-th sub-historical period becomes the second user in the (i+1)-th sub-historical period, calculate the third probability that the second user in the i-th sub-historical period becomes the first user in the (i+1)-th sub-historical period, and calculate the fourth probability that the second user in the i-th sub-historical period becomes the second user in the (i+1)-th sub-historical period, where i is a positive integer; Based on the first probability, the second probability, the third probability, and the fourth probability, the first transformation probability matrix corresponding to the first discrete feature is obtained; The second discrete feature corresponding to the first discrete feature is determined based on the first transformation probability matrix corresponding to the first discrete feature.

3. The method according to claim 1, characterized in that, Before classifying multiple users into multiple user groups based on target classification theory and target classification models, the following is also included: Obtain the characteristics of the multiple users in the historical period; Based on feature engineering, the features of the multiple users in the historical period are processed to obtain the features of the multiple users after feature engineering. The acquisition of the first target features of N user groups in a historical period includes: Based on the features processed by the multiple user features, the first target features of the N user groups in the historical period are obtained.

4. The method according to claim 3, characterized in that, Before processing the features of the multiple users over historical periods based on feature engineering to obtain the feature-engineered features of the multiple users, the method further includes: Data cleaning is performed on the characteristics of the multiple users in historical periods to obtain the cleaned characteristics of the multiple user data. The features of the multiple user data after cleaning are normalized to obtain the normalized features of the multiple users. The feature engineering process is used to process the features of the multiple users over a historical period to obtain the feature-engineered features of the multiple users, including: Based on feature engineering, the normalized features of the multiple users are processed to obtain the feature-engineered features of the multiple users.

5. The method according to claim 1, characterized in that, The step of determining the corresponding phone plan for each user group based on the second target features of each user group in the future period and at least one machine learning model includes: For any user group's second target feature in the future period, the user group's second target feature in the future period is input into multiple different machine learning models to obtain the prediction results of multiple machine learning models for the corresponding call charges of the user group; When the prediction results of a preset number of machine learning models are consistent, the prediction results of the preset number of machine learning models are used as the phone bill packages corresponding to the user group.

6. The method according to claim 5, characterized in that, The various machine learning models include at least two of the following: random forest model, backpropagation neural network model, support vector machine model, logistic regression model, distributed gradient boosting model, and K-nearest neighbor classification algorithm model.

7. A device for determining a package, characterized in that, include: The classification module is used to divide multiple users into multiple user groups according to requirements based on target classification theory and target classification model. The target classification theory includes survival, interrelationship and growth theories. For any user group, based on at least one of the following, a clustering algorithm is used to further subdivide the user group into at least one user group. The user groups in the multiple user groups are then aggregated to obtain N user groups. The first acquisition module is used to acquire the first target features of N user groups in a historical period. The first target features include x first discrete features and y first continuous features, where N is a positive integer and x and y are both positive integers. The prediction module is used to obtain x second discrete features of the user group in future periods based on the target transition probability algorithm and the x first discrete features of the user group in historical periods for any given user group. Based on the target transition probability algorithm and the y first continuous features of the user group in the historical period, the y second continuous features of the user group in the future period are obtained. The first determining module is used to determine the corresponding call charge package for each user group based on the second target features of each user group in the future period and at least one machine learning model, wherein the second target features include the x second discrete features and the y second continuous features.

8. An electronic device, characterized in that, The electronic device includes: a processor, a memory, and a computer program stored in the memory and executable on the processor, wherein the computer program, when executed by the processor, implements the steps of the method for determining the package as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, A computer program is stored on the computer-readable storage medium, which, when executed by a processor, implements the steps of the method for determining the package as described in any one of claims 1 to 6.

Images