A recommended method, apparatus, electronic device, and storage medium

Abstract

This invention provides a recommendation method, apparatus, electronic device, and readable storage medium. The method includes: acquiring communication package information corresponding to each communication package, and acquiring basic user information and communication data information of the user to be recommended; determining basic user features based on the basic user information, determining communication data features based on the communication data information, and determining communication package features based on the communication package information; inputting the basic user features into a pre-trained neural network model, and outputting the selection probability of the user to be recommended for each communication package; calculating the similarity between the communication data features and the features of each communication package using a preset correlation coefficient algorithm; and determining the target communication package to be recommended to the user based on the selection probability and similarity. This recommendation method can improve the accuracy of recommended packages and solve the problem that traditional recommendation algorithms based on association rule mining cannot recommend packages to new users in the prior art.

Description

Technical Field

[0001] This invention relates to the field of computer technology, and in particular to a recommendation method, a recommendation device, an electronic device, and a computer-readable storage medium. Background Technology

[0002] When providing online subscription services, users are typically categorized based on their current subscription plans and benefits, and recommendations are then made to users with similar preferences. However, traditional recommendation algorithms often end up suggesting plans that users are not particularly interested in. Furthermore, for new users (those without prior subscription plans or benefits), traditional algorithms cannot effectively recommend suitable plans because there is no existing record of these features. Summary of the Invention

[0003] In view of the above problems, embodiments of the present invention are proposed to provide a recommended method and a corresponding recommended apparatus, an electronic device, and a computer-readable storage medium to overcome or at least partially solve the above problems.

[0004] This invention discloses a recommended method, the method comprising:

[0005] Obtain the communication package information corresponding to each communication package, as well as the basic user information and communication data information of the users to be recommended;

[0006] Determine the basic user characteristics for the user's basic information, determine the communication data characteristics for the communication data information, and determine the communication package characteristics for the communication package information respectively;

[0007] The user's basic characteristics are input into a pre-trained neural network model, which outputs the probability of the user to be recommended choosing each of the communication packages.

[0008] The similarity between the communication data features and the features of each of the communication packages is calculated using a preset correlation coefficient algorithm;

[0009] The target communication package to be recommended to the user is determined based on the selection probability and the similarity.

[0010] Optionally, determining the target communication package to recommend to the user based on the selection probability and the similarity includes:

[0011] For any of the communication packages, the selection probability and similarity corresponding to the communication package are weighted and averaged to obtain the weighted average value corresponding to the communication package.

[0012] The maximum value is determined from the weighted average values ​​corresponding to each of the aforementioned communication packages, and the communication package corresponding to the maximum value is determined as the target communication package.

[0013] Optionally, the pre-trained neural network model is trained in the following manner:

[0014] Identify the sample users to be used as training samples, determine the basic information of the sample users, and determine the target communication packages that the sample users have subscribed to.

[0015] Extract the basic characteristics of the sample users from the basic information of the sample users;

[0016] The model is trained using the basic features of the sample users and the target communication package of the sample to obtain the model weight parameters of the neural network model; the model weight parameters include a matrix vector and an offset vector.

[0017] Optionally, determining the sample users as training samples includes:

[0018] Acquire candidate sample users, determine the candidate sample communication data information of the candidate sample users, and determine the candidate sample target communication package information of the candidate sample target communication packages that the candidate sample users have subscribed to;

[0019] For any candidate sample user, a correlation analysis is performed on the candidate sample communication data information and the candidate sample target communication package information corresponding to the candidate sample user to obtain the communication data correlation value corresponding to the candidate sample user.

[0020] The communication data correlation value corresponding to each of the candidate sample users is determined, and the candidate sample users whose communication data correlation value is greater than a preset correlation threshold are determined as the sample users as training samples.

[0021] Optionally, the communication data features include voice data features, data traffic features, SMS data features, and value-added service data features. The step of calculating the similarity between the communication data features and the features of each communication package using a preset correlation coefficient algorithm includes:

[0022] Determine the target age group to which the user to be recommended belongs, and determine the weight coefficients to be assigned to the voice data features, traffic data features, SMS data features and value-added service data features in the target age group respectively;

[0023] The preset correlation coefficient algorithm is determined based on the weight coefficient, and the similarity between the communication data features and the features of each communication package is calculated using the preset correlation coefficient algorithm.

[0024] Optionally, the algorithm for determining the preset correlation coefficient based on the weighting coefficient includes:

[0025] The Pearson correlation coefficient algorithm is improved based on the weighting coefficients, and the improved Pearson correlation coefficient algorithm is determined as the preset correlation coefficient algorithm.

[0026] Optionally, determining the user basic characteristics for the user basic information, determining the communication data characteristics for the communication data information, and determining the communication package characteristics for the communication package information respectively include:

[0027] The user's basic information is subjected to word embedding processing to obtain the corresponding user basic features;

[0028] The communication data information is subjected to deviation standardization processing to obtain the corresponding communication data features;

[0029] The communication package information is subjected to deviation standardization processing to obtain the corresponding communication package features.

[0030] This invention also discloses a recommended device, the device comprising:

[0031] The acquisition module is used to acquire communication package information corresponding to each communication package, as well as basic user information and communication data information of the user to be recommended;

[0032] The first determining module is used to determine the user basic characteristics for the user basic information, the communication data characteristics for the communication data information, and the communication package characteristics for the communication package information respectively.

[0033] The input / output module is used to input the user's basic features into a pre-trained neural network model and output the probability of the user to be recommended choosing each of the communication packages.

[0034] The calculation module is used to calculate the similarity between the communication data features and the features of each of the communication packages using a preset correlation coefficient algorithm;

[0035] The second determining module is used to determine the target communication package to recommend to the user to be recommended based on the selection probability and the similarity.

[0036] Optionally, the second determining module includes:

[0037] The weighted average calculation submodule is used to calculate the weighted average of the selection probability and similarity corresponding to any one of the communication packages, and obtain the weighted average value corresponding to the communication package.

[0038] The first determining submodule is used to determine the maximum value from the weighted average values ​​corresponding to each of the communication packages, and to determine the communication package corresponding to the maximum value as the target communication package.

[0039] Optionally, the apparatus is further configured to train and obtain the neural network model, and the apparatus further includes:

[0040] The third determining module is used to determine the sample users as training samples, determine the basic information of the sample users, and determine the sample target communication package that the sample users have subscribed to.

[0041] The extraction module is used to extract the basic features of the sample users from the basic information of the sample users.

[0042] The model training module is used to train the model using the basic features of the sample users and the target communication package of the sample, and to obtain the model weight parameters of the neural network model; the model weight parameters include a matrix vector and an offset vector.

[0043] Optionally, the third determining module includes:

[0044] The acquisition and determination submodule is used to acquire candidate sample users, determine the candidate sample communication data information of the candidate sample users, and determine the candidate sample target communication package information of the candidate sample users who have subscribed to the candidate sample target communication package.

[0045] The correlation analysis submodule is used to perform correlation analysis on the candidate sample communication data information and the candidate sample target communication package information corresponding to any candidate sample user, and to obtain the communication data correlation value corresponding to the candidate sample user.

[0046] The second determining submodule is used to determine the communication data correlation value corresponding to each of the candidate sample users, and to determine the candidate sample users whose communication data correlation value is greater than a preset correlation threshold as the sample users used as training samples.

[0047] Optionally, the communication data features include voice data features, traffic data features, SMS data features, and value-added service data features, and the calculation module includes:

[0048] The third determining submodule is used to determine the target age group to which the age of the user to be recommended belongs, and to determine the weight coefficients to be assigned to the voice data features, the traffic data features, the SMS data features and the value-added service data features in the target age group respectively;

[0049] The determination and calculation submodule is used to determine the preset correlation coefficient algorithm based on the weight coefficient, and to use the preset correlation coefficient algorithm to calculate the similarity between the communication data features and the features of each communication package.

[0050] Optionally, the third determining submodule includes:

[0051] The Pearson correlation coefficient algorithm is improved based on the weighting coefficients, and the improved Pearson correlation coefficient algorithm is determined as the preset correlation coefficient algorithm.

[0052] Optionally, the first determining module includes:

[0053] The word embedding processing submodule is used to perform word embedding processing on the user's basic information to obtain the corresponding user basic features;

[0054] The first deviation standardization processing submodule is used to perform deviation standardization processing on the communication data information to obtain the corresponding communication data features;

[0055] The second deviation standardization processing submodule is used to perform deviation standardization processing on the communication package information to obtain the corresponding communication package features.

[0056] This invention also discloses an electronic device, including: a processor, a memory, and a computer program stored in the memory and capable of running on the processor, wherein the computer program, when executed by the processor, implements the steps of a recommended method as described above.

[0057] This invention also discloses a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of a recommended method as described above.

[0058] The embodiments of the present invention have the following advantages:

[0059] In this embodiment of the invention, communication package information corresponding to each communication package can be obtained, as well as basic user information and communication data information of the user to be recommended. Basic user features based on the basic user information, communication data features based on the communication data information, and communication package features based on the communication package information can be extracted. The basic user features can be input into a pre-trained neural network model, which outputs the probability of the user selecting each communication package. A preset correlation coefficient algorithm is used to calculate the similarity between the communication data features and the features of each communication package. Based on the selection probability and similarity, the target communication package to be recommended to the user is determined. By employing the above method, using neural network technology, the probability of the current user selecting each communication package is predicted using the current user's basic user information data, thus constructing a neural network-based package recommendation model. Based on the correlation coefficient in statistics, a preset correlation coefficient algorithm is used to calculate the similarity between the current user's communication data information and the information of each communication package, thus constructing a statistically based package recommendation model. The target communication package can be determined jointly based on the prediction results of the above two models and recommended to the current user. This recommendation method can improve the accuracy of package recommendations and solve the problem that traditional recommendation algorithms based on association rule mining in the prior art cannot recommend packages to new users. Attached Figure Description

[0060] Figure 1 This is a flowchart of the steps of a recommended method provided in an embodiment of the present invention;

[0061] Figure 2 This is a flowchart of another recommended method provided in an embodiment of the present invention;

[0062] Figure 3 This is a schematic diagram illustrating an application scenario of a recommended method provided in an embodiment of the present invention;

[0063] Figure 4 This is a structural block diagram of a recommended device provided in an embodiment of the present invention. Detailed Implementation

[0064] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art are within the scope of protection of the present invention.

[0065] When providing online subscription services, users are typically categorized based on their current subscription plans and benefits, and recommendations are then made to users with similar preferences. However, traditional recommendation algorithms often end up suggesting plans that users are not particularly interested in. Furthermore, for new users (those without prior subscription plans or benefits), traditional algorithms cannot effectively recommend suitable plans because there is no existing record of these features.

[0066] Therefore, the present invention aims to provide a recommended method and a corresponding recommended device, an electronic device, and a computer-readable storage medium to overcome or at least partially solve the above problems.

[0067] One of the core concepts of this invention is the ability to obtain communication package information corresponding to each communication package, as well as the user's basic information and communication data information of the user to be recommended. This involves extracting basic user features based on the user's basic information, communication data features based on the communication data information, and communication package features based on the communication package information. The user's basic features can be input into a pre-trained neural network model, which outputs the probability of the user selecting each communication package. A preset correlation coefficient algorithm is used to calculate the similarity between the communication data features and the features of each communication package. Based on the selection probability and similarity, the target communication package to be recommended to the user is determined. By employing the above method, using neural network technology, the probability of the current user selecting each communication package is predicted using the current user's basic information data, thus constructing a neural network-based package recommendation model. Based on the correlation coefficient in statistics, a preset correlation coefficient algorithm is used to calculate the similarity between the current user's communication data information and the information of each communication package, thus constructing a statistically based package recommendation model. The target communication package can be determined jointly based on the prediction results of the two models and recommended to the current user. This recommendation method can improve the accuracy of package recommendations and solve the problem that traditional recommendation algorithms based on association rule mining cannot recommend packages to new users.

[0068] Reference Figure 1 The diagram illustrates a flowchart of a recommended method provided by an embodiment of the present invention, which may specifically include the following steps:

[0069] Step 101: Obtain the communication package information corresponding to each communication package, as well as the user basic information and communication data information of the user to be recommended.

[0070] In this embodiment of the invention, communication package information corresponding to each communication package can be obtained. Communication package information refers to the package content included in the communication packages currently offered by the operator, which may include voice call limits, data limits, SMS limits, and value-added service fees. For example, assuming three communication packages are currently offered, namely communication package A, communication package B, and communication package C, their corresponding communication package information can be:

[0071] Communication Package A: Voice call limit: 150 minutes / month, Data limit: 15GB / month, SMS limit: 100 messages / month, Value-added service fee: 15 yuan / month;

[0072] Communication Package B: Voice call limit: 200 minutes / month, Data limit: 12GB / month, SMS limit: 50 messages / month, Value-added service fee: 13 yuan / month;

[0073] Communication Package C: Voice call limit: 400 minutes / month, Data limit: 20GB / month, SMS limit: 150 messages / month, Value-added service fee: 10 yuan / month.

[0074] Identify the users who need to be recommended communication packages, and obtain their basic user information and communication data.

[0075] User basic information includes one or more of the following: age, gender, education, occupation, income, mobile number registration location, mobile number network tenure, and customer rating.

[0076] Communication data information can refer to communication data usage, which may include voice usage, data usage, SMS usage, and value-added service usage. For example, communication data information can refer to a user's recent average monthly usage, including: voice usage (min / month), data usage (GB / month), SMS usage (messages / month), and value-added service usage (yuan / month).

[0077] For example, it can obtain the basic user information of the user to be recommended, the average communication data over the past 3 months, and the information on all current communication plans. The information collected is in the following formats:

[0078] The basic information of the user to be recommended is as follows: Age: 26, Gender: Male, Education: Bachelor's Degree, Occupation: Teacher, Income: 10, Mobile Number Registration Location: Suzhou City, Mobile Number Tenure: 2, Customer Rating: 3;

[0079] The average communication data for the past 3 months is as follows: voice usage: 190 minutes / month, data usage: 10GB / month, SMS usage: 63 messages / month, and value-added service usage: 15 yuan / month.

[0080] Suppose that three communication packages are currently offered:

[0081] Communication Package A: Voice call limit: 150 minutes / month, Data limit: 15GB / month, SMS limit: 100 messages / month, Value-added service fee: 15 yuan / month;

[0082] Communication Package B: Voice call limit: 200 minutes / month, Data limit: 12GB / month, SMS limit: 50 messages / month, Value-added service fee: 13 yuan / month;

[0083] Communication Package C: Voice call limit: 400 minutes / month, Data limit: 20GB / month, SMS limit: 150 messages / month, Value-added service fee: 10 yuan / month.

[0084] Step 102: Determine the basic characteristics of the user for the basic user information, determine the communication data characteristics for the communication data information, and determine the communication package characteristics for the communication package information respectively.

[0085] Feature extraction is performed on user basic information to obtain corresponding user basic features; feature extraction is performed on communication data information to obtain corresponding communication data features; feature extraction is performed on each communication package information to obtain corresponding communication package features.

[0086] Step 103: Input the user's basic features into the pre-trained neural network model and output the probability of the user's choice of each communication package.

[0087] The user's basic characteristics are used as input to a pre-trained neural network model, which outputs the probability of the user's choice of each communication package, forming a corresponding probability list.

[0088] Step 104: Calculate the similarity between the communication data features and the features of each communication package using a preset correlation coefficient algorithm.

[0089] A pre-defined correlation coefficient algorithm is used to calculate the similarity between communication data features and the features of each communication package, forming a corresponding similarity list.

[0090] Step 105: Determine the target communication package to recommend to the user based on the selection probability and similarity.

[0091] The target communication package to be recommended to the user is determined by combining the selection probability in the probability list and the similarity in the similarity list.

[0092] In summary, in this embodiment of the invention, a neural network-based plan recommendation model is constructed by using neural network technology to predict the probability of the current user selecting various communication plans based on the current user's basic information data. A statistically based plan recommendation model is also constructed by using a preset correlation coefficient algorithm to calculate the similarity between the current user's communication data and the information of each communication plan. The target communication plan can be determined based on the prediction results of the two models and recommended to the current user. This recommendation method can improve the accuracy of plan recommendations and solve the problem that existing recommendation algorithms based on traditional association rule mining cannot recommend plans to new users.

[0093] Reference Figure 2 The diagram illustrates a flowchart of another recommended method provided by an embodiment of the present invention, which may specifically include the following steps:

[0094] Step 201: Obtain the communication package information corresponding to each communication package, as well as the user basic information and communication data information of the user to be recommended.

[0095] In this embodiment of the invention, communication package information corresponding to each communication package can be obtained, and users to be recommended for communication package recommendations can be identified. The basic user information and communication data information of the users to be recommended can be obtained. For example, web crawling technology can be used to obtain the basic user information, communication data information, and communication package information of all current communication packages for target users applying for network access services online.

[0096] Step 202: Determine the basic characteristics of the user for the basic user information, determine the communication data characteristics for the communication data information, and determine the communication package characteristics for the communication package information respectively.

[0097] In an optional embodiment of the present invention, step 202, which determines the user's basic characteristics for basic user information, the communication data characteristics for communication data information, and the communication package characteristics for communication package information, may specifically include the following sub-steps:

[0098] Sub-step S11 involves word embedding processing of the user's basic information to obtain the corresponding user basic features.

[0099] In this embodiment of the invention, word embedding processing can be performed on user basic information to obtain corresponding user basic features. To enable the Chinese character category features in the user basic information to be input into the neural network model, word embedding is used to process the Chinese character category features of the user basic information to obtain the user basic features, i.e., the user basic feature vector, as shown below:

[0100] X basic =[xb1 ,x b2 ,x b3 ,x b4 ,x b5 ,x b6 ,x b7 ,x b8 ] T ;

[0101] Among them, X basic Represents the user's basic feature vector; x b1 Represents the age feature vector; x b2 Represents the gender feature vector; x b3 Represents the educational background feature vector; x b4 Represents the occupational feature vector; x b5 Represents the income eigenvector; x b6 Represents the feature vector of the mobile number's registration location; x b7 Represents the mobile number's network age feature vector; x b8 This represents the feature vector of customer star rating.

[0102] Sub-step S12 involves performing deviation standardization on the communication data information to obtain the corresponding communication data features.

[0103] Sub-step S13: Perform deviation standardization processing on the communication package information to obtain the corresponding communication package features.

[0104] In this embodiment of the invention, deviation standardization processing can be performed on communication data information and communication package information to obtain corresponding communication data features and communication package features.

[0105] To eliminate the influence of dimensions on the model, the communication data information and communication package information are standardized according to the deviation standardization formula to obtain the standardized communication data features and communication package features, namely the communication data feature vector and the communication package feature vector. The deviation standardization formula is shown in formula (1) below:

[0106]

[0107] Where x max x represents the maximum value of the original sample data. min Let x be the minimum value of the original sample data, and let x be the sample value to be processed. * This represents the sample value after standard deviation processing.

[0108] Taking the data from step 101 as an example, the resulting communication data feature vector is X. actual =[x a1 ,x a2 ,x a3 ,x a4 ]T The feature vector of communication package A is X. A =[x A1 ,x A2 ,x A3 ,x A4 ] T The feature vector of communication package B is X. B =[x B1 ,x B2 ,x B3 ,x B4 ] T The feature vector of communication package C is X. C =[x C1 ,x C2 ,x C3 ,x C4 ] T .

[0109] Step 203: Input the user's basic features into the pre-trained neural network model and output the probability of the user's choice of each communication package.

[0110] A pre-trained neural network model is used for probabilistic prediction. Before predicting the results based on the neural network model, the model needs to be trained using training samples from the training set. For example, the input is a user feature vector X composed of the user's basic information to be recommended. basic By making predictions, we obtain a list of probabilities P = [p1, p2, p3] for the current user to be recommended to choose each communication package. T .

[0111] In an optional embodiment of the present invention, the pre-trained neural network model is trained in the following manner:

[0112] The sample users to be used as training samples are identified, along with their basic information and the target communication packages they have subscribed to. The basic features of the sample users are extracted from their basic information. The model is then trained using these basic features and the target communication packages to obtain the model weight parameters of the neural network model.

[0113] The model weight parameters include a matrix vector and an offset vector.

[0114] In this embodiment of the invention, sample users as training samples can be determined, and the basic information of the sample users and the target communication packages that the sample users have subscribed to can be determined. The basic features of the sample users are extracted from the basic information of the sample users, and the basic features of the sample users and the target communication packages are used to train the model and obtain the model weight parameters of the neural network model.

[0115] Specifically, after determining the training samples, the neural network model is trained using the training samples to obtain the trained model weight parameters, including the matrix vector W. (i) and offset vector b (i) W (i) Specifically b (i) Specifically K is the number of hidden layers.

[0116] In an optional embodiment of the present invention, the activation function in the neural network model can be the Sigmoid function. Since the present invention needs to calculate the probability of a user choosing each communication package, the activation function chosen is the Sigmoid function, which is convenient for calculating probabilities. The Sigmoid function is a common S-shaped function in biology, also known as an S-shaped growth curve. In information science, due to its monotonically increasing and inversely increasing properties, the Sigmoid function is often used as the activation function of neural networks to map variables to the range of 0 and 1. The Sigmoid function calculation formula can be shown in the following formula (2):

[0117]

[0118] Let the intermediate hidden layer vector be Y. (1) ,Y (2) ,…,Y (K) The hidden layer 1, hidden layer K, and output vector P calculated based on the input vector and weights can be represented by the following formulas (3) to (5):

[0119]

[0120]

[0121] P = f(W) (K+1) Y (K) +b (K+1) = [p1, p2, p3] T (5)

[0122] In an optional embodiment of the present invention, determining the sample users as training samples may specifically include the following sub-steps:

[0123] Sub-step S21: Obtain candidate sample users, determine candidate sample communication data information of candidate sample users, and determine candidate sample target communication package information of candidate sample users who have subscribed to candidate sample target communication packages.

[0124] Sub-step S22: For any candidate sample user, perform correlation analysis on the candidate sample communication data information and the candidate sample target communication package information corresponding to the candidate sample user to obtain the communication data correlation value corresponding to the candidate sample user.

[0125] Sub-step S23: Determine the communication data correlation value corresponding to each candidate sample user, and determine the candidate sample users whose communication data correlation value is greater than the preset correlation threshold as the sample users used as training samples.

[0126] Candidate sample users are the candidate sample users. We can collect the candidate sample communication data information of the candidate sample users and the candidate sample target communication package information of the candidate sample users selected by the candidate sample users. Then, we perform correlation analysis on the candidate sample communication data information and the selected candidate sample target communication package information of each candidate sample user, and retain the candidate sample users with high correlation as training samples.

[0127] Step 204: Calculate the similarity between the communication data features and the features of each communication package using a preset correlation coefficient algorithm.

[0128] In this embodiment of the invention, similarity is calculated based on a statistical model.

[0129] In an optional embodiment of the present invention, the communication data features include voice data features, traffic data features, SMS data features, and value-added service data features. Step 204 uses a preset correlation coefficient algorithm to calculate the similarity between the communication data features and the features of each communication package, which may specifically include the following sub-steps:

[0130] Sub-step S31: Determine the target age group to which the user to be recommended belongs, and determine the weight coefficients to be assigned to voice data features, traffic data features, SMS data features and value-added service data features in the target age group.

[0131] The weighting coefficients for each age group can be calculated using quantitative statistical methods. First, the age group is divided into six intervals: under 18, 18-27, 28-37, 38-47, 48-57, and over 57. Then, sufficient questionnaires are distributed to users in each age group to collect data on the importance of voice calls, data traffic, SMS, and value-added services. Assuming the statistical results for the 18-27 age group are shown in Table 1, the process of calculating the weighting coefficients based on the questionnaire results is as follows:

[0132] First, assign 0 to unimportant, 1 to important, and 2 to very important. Then, the weights of these three options are as follows:

[0133]

[0134] Table 1. Statistical Results of Importance Evaluation of Voice, Data, SMS, and Value-Added Services by Age Groups of 18-27

[0135] Second, calculate the weight of each indicator. Taking the voice indicator as an example, the calculation is shown in formula (6).

[0136]

[0137] Based on the above formula, a weighted coefficient vector w = [w1, w2, w3, w4] can be obtained for each age group. T , respectively, represent the weighting coefficients to be assigned to voice, data traffic, SMS, and value-added services for this age group, that is, the weighting coefficients to be assigned to voice data features, data traffic features, SMS data features, and value-added service data features for this age group.

[0138] Sub-step S32: Determine the preset correlation coefficient algorithm based on the weight coefficient, and use the preset correlation coefficient algorithm to calculate the similarity between the communication data features and the features of each communication package.

[0139] In this embodiment of the invention, a preset correlation coefficient algorithm can be determined based on the weighting coefficient.

[0140] In an optional embodiment of the present invention, the algorithm for determining the preset correlation coefficient based on the weight coefficient in sub-step S22 may specifically include the following sub-steps:

[0141] The Pearson correlation coefficient algorithm is improved based on the weighting coefficient, and the improved Pearson correlation coefficient algorithm is determined as the preset correlation coefficient algorithm.

[0142] The traditional Pearson correlation coefficient algorithm, as shown in formula (7), has some problems. For example, it only considers the numerical correlation of the data itself, and assumes that each data point contributes equally to the calculation of the correlation coefficient, simply assigning them the same weight, while ignoring the different contributions of different data points to the calculation of the correlation coefficient. In this invention, users of different ages have different needs for voice, data, SMS, and value-added services in the package. For example, users aged 18-27 mostly value data and value-added services more, and when calculating the correlation coefficient, data and value-added services have a greater impact on the correlation, i.e., a larger weight; while users aged 48-57 may value voice and SMS more, and when calculating similarity, voice and SMS have a larger weight. Therefore, this invention improves the traditional Pearson correlation coefficient by adding an age-based weight coefficient in the correlation coefficient calculation. The improved Pearson correlation coefficient algorithm formulas are shown in (8) to (9):

[0143]

[0144]

[0145]

[0146] Where, x i ,y i For sample data values, w i represents the weighting coefficients for a certain age group in voice, data, SMS, and value-added services, and m(x,w) is the weighted average calculated according to the weighting coefficients.

[0147] After performing deviation standardization on the communication data and communication package information, the feature vector X of the current user's communication data to be recommended is calculated according to the improved Pearson correlation coefficient algorithm formula. actual The feature vector X of each current communication package A X B X C The correlation coefficient is used to obtain a similarity list γ = [γ1, γ2, γ3] for the current users to be recommended. T For example, X actual With one of the communication package feature vectors X A Based on the weight coefficient vector w = [w1, w2, w3, w4] T The formula for calculating the correlation coefficient is:

[0148]

[0149] Step 205: For any communication package, calculate the weighted average of the selection probability and similarity corresponding to that communication package to obtain the weighted average value corresponding to that communication package.

[0150] Step 206: Determine the maximum value from the weighted average values ​​corresponding to each communication package, and determine the communication package corresponding to the maximum value as the target communication package.

[0151] The obtained probability list and similarity list are weighted and averaged, then arranged in ascending order, and the communication package corresponding to the maximum value is recommended to the current user.

[0152] In an optional embodiment of the present invention, the model weight parameters in the neural network model can also be adjusted according to the recommendation result, wherein the recommendation result refers to whether the user to be recommended chooses to adopt the target communication package recommended this time.

[0153] This process is mainly divided into three stages: sample collection, query, and training. In the sample collection stage, the model collects new samples based on whether the user selects the recommended package. If the user rejects the recommended package, the recommendation process ends; if the user selects the recommended package, the user's basic information and the selected package are actively added to the training set as new samples. In the query stage, the model periodically queries the training set for the number of newly added samples. The query stage ends when the number of new samples reaches a set threshold. Next, the training stage begins. In this stage, the new samples are used to fine-tune the neural network model. After the required accuracy is achieved, the fine-tuned model weight parameters are obtained, and the fine-tuned model is then used for prediction.

[0154] Reference Figure 3 The diagram illustrates an application scenario of a recommendation method provided by an embodiment of the present invention. The recommendation method provided by the present invention can be applied to scenarios where communication packages are recommended to users. In specific implementation, the recommendation platform can obtain communication package information corresponding to each communication package, as well as the user's basic information and communication data information. It can extract corresponding user basic features, communication data features, and communication package features. Then, the user's basic features can be input into a pre-trained neural network model to determine the user's selection probability for each communication package. Furthermore, a preset correlation coefficient algorithm can be used to calculate the similarity between the user's communication data features and the features of each communication package. Based on the selection probability and similarity, a target communication package is determined to recommend to the user, and then the target communication package is pushed to the corresponding user terminal. The user terminal can be a mobile terminal, such as a smartphone. The recommendation platform can integrate one or more servers or be part of a cloud computing resource pool.

[0155] In summary, in this embodiment of the invention, a neural network-based plan recommendation model is constructed by using neural network technology to predict the probability of the current user selecting various communication plans based on the current user's basic information data. A statistically based plan recommendation model is also constructed by using a preset correlation coefficient algorithm to calculate the similarity between the current user's communication data and the information of each communication plan. The target communication plan can be determined based on the prediction results of the two models and recommended to the current user. This recommendation method can improve the accuracy of plan recommendations and solve the problem that existing recommendation algorithms based on traditional association rule mining cannot recommend plans to new users.

[0156] It should be noted that, for the sake of simplicity, the method embodiments are all described as a series of actions. However, those skilled in the art should understand that the embodiments of the present invention are not limited to the described order of actions, because according to the embodiments of the present invention, some steps can be performed in other orders or simultaneously. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions involved are not necessarily essential to the embodiments of the present invention.

[0157] Reference Figure 4 The diagram shows a structural block diagram of a recommended device provided by an embodiment of the present invention, which may specifically include the following modules:

[0158] The acquisition module 401 is used to acquire communication package information corresponding to each communication package, as well as user basic information and communication data information of the user to be recommended;

[0159] The first determining module 402 is used to determine the user basic characteristics for the user basic information, the communication data characteristics for the communication data information, and the communication package characteristics for the communication package information respectively.

[0160] The input / output module 403 is used to input the user's basic features into a pre-trained neural network model and output the probability of the user to be recommended choosing each of the communication packages.

[0161] The calculation module 404 is used to calculate the similarity between the communication data features and the features of each of the communication packages using a preset correlation coefficient algorithm;

[0162] The second determining module 405 is used to determine the target communication package to recommend to the user to be recommended based on the selection probability and the similarity.

[0163] In this embodiment of the invention, the second determining module includes:

[0164] The weighted average calculation submodule is used to calculate the weighted average of the selection probability and similarity corresponding to any one of the communication packages, and obtain the weighted average value corresponding to the communication package.

[0165] The first determining submodule is used to determine the maximum value from the weighted average values ​​corresponding to each of the communication packages, and to determine the communication package corresponding to the maximum value as the target communication package.

[0166] In this embodiment of the invention, the apparatus is further configured to train and obtain the neural network model, and the apparatus further includes:

[0167] The third determining module is used to determine the sample users as training samples, determine the basic information of the sample users, and determine the sample target communication package that the sample users have subscribed to.

[0168] The extraction module is used to extract the basic features of the sample users from the basic information of the sample users.

[0169] The model training module is used to train the model using the basic features of the sample users and the target communication package of the sample, and to obtain the model weight parameters of the neural network model; the model weight parameters include a matrix vector and an offset vector.

[0170] In this embodiment of the invention, the third determining module includes:

[0171] The acquisition and determination submodule is used to acquire candidate sample users, determine the candidate sample communication data information of the candidate sample users, and determine the candidate sample target communication package information of the candidate sample users who have subscribed to the candidate sample target communication package.

[0172] The correlation analysis submodule is used to perform correlation analysis on the candidate sample communication data information and the candidate sample target communication package information corresponding to any candidate sample user, and to obtain the communication data correlation value corresponding to the candidate sample user.

[0173] The second determining submodule is used to determine the communication data correlation value corresponding to each of the candidate sample users, and to determine the candidate sample users whose communication data correlation value is greater than a preset correlation threshold as the sample users used as training samples.

[0174] In this embodiment of the invention, the communication data features include voice data features, traffic data features, SMS data features, and value-added service data features; the calculation module includes:

[0175] The third determining submodule is used to determine the target age group to which the age of the user to be recommended belongs, and to determine the weight coefficients to be assigned to the voice data features, the traffic data features, the SMS data features and the value-added service data features in the target age group respectively;

[0176] The determination and calculation submodule is used to determine the preset correlation coefficient algorithm based on the weight coefficient, and to use the preset correlation coefficient algorithm to calculate the similarity between the communication data features and the features of each communication package.

[0177] In this embodiment of the invention, the third determining submodule includes:

[0178] The Pearson correlation coefficient algorithm is improved based on the weighting coefficients, and the improved Pearson correlation coefficient algorithm is determined as the preset correlation coefficient algorithm.

[0179] In this embodiment of the invention, the first determining module includes:

[0180] The word embedding processing submodule is used to perform word embedding processing on the user's basic information to obtain the corresponding user basic features;

[0181] The first deviation standardization processing submodule is used to perform deviation standardization processing on the communication data information to obtain the corresponding communication data features;

[0182] The second deviation standardization processing submodule is used to perform deviation standardization processing on the communication package information to obtain the corresponding communication package features.

[0183] In summary, in this embodiment of the invention, a neural network-based plan recommendation model is constructed by using neural network technology to predict the probability of the current user selecting various communication plans based on the current user's basic information data. A statistically based plan recommendation model is also constructed by using a preset correlation coefficient algorithm to calculate the similarity between the current user's communication data and the information of each communication plan. The target communication plan can be determined based on the prediction results of the two models and recommended to the current user. This recommendation method can improve the accuracy of plan recommendations and solve the problem that existing recommendation algorithms based on traditional association rule mining cannot recommend plans to new users.

[0184] As the device embodiment is basically similar to the method embodiment, the description is relatively simple, and relevant parts can be found in the description of the method embodiment.

[0185] This invention also provides an electronic device, including: a processor, a memory, and a computer program stored in the memory and capable of running on the processor. When the computer program is executed by the processor, it implements the various processes of the above-described recommended method embodiment and achieves the same technical effect. To avoid repetition, it will not be described again here.

[0186] This invention also provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, it implements the various processes of the above-described recommended method embodiment and achieves the same technical effect. To avoid repetition, it will not be described again here.

[0187] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0188] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, apparatus, or computer program products. Therefore, embodiments of the present invention can take the form of entirely hardware embodiments, entirely software embodiments, or embodiments combining software and hardware aspects. Furthermore, embodiments of the present invention can take the form of computer program products implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0189] Embodiments of the present invention are described with reference to flowchart illustrations and / or block diagrams of methods, terminal devices (systems), and computer program products according to embodiments of the invention. It will be understood that each block of 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, special-purpose computer, embedded processor, or other programmable data processing terminal device to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing terminal device, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0190] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing terminal device to operate in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0191] These computer program instructions can also be loaded onto a computer or other programmable data processing terminal equipment, causing a series of operational steps to be performed on the computer or other programmable terminal equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable terminal equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0192] Although preferred embodiments of the present invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of the embodiments of the present invention.

[0193] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only 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 terminal device 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 terminal device. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or terminal device that includes said element.

[0194] The foregoing has provided a detailed description of a recommended method, a recommended device, an electronic device, and a computer-readable storage medium provided by the present invention. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A recommendation method, characterized in that, The method includes: Obtain the communication package information corresponding to each communication package, as well as the basic user information and communication data information of the users to be recommended; Determine the basic user characteristics for the user's basic information, determine the communication data characteristics for the communication data information, and determine the communication package characteristics for the communication package information respectively; The user's basic characteristics are input into a pre-trained neural network model, which outputs the probability of the user to be recommended choosing each of the communication packages. The similarity between the communication data features and the features of each of the communication packages is calculated using a preset correlation coefficient algorithm; The target communication package to be recommended to the user is determined based on the selection probability and the similarity. The communication data features include voice data features, data traffic features, SMS data features, and value-added service data features. The calculation of the similarity between the communication data features and the features of each communication package using a preset correlation coefficient algorithm includes: Determine the target age group to which the user to be recommended belongs, and determine the weight coefficients to be assigned to the voice data features, traffic data features, SMS data features and value-added service data features in the target age group respectively; The preset correlation coefficient algorithm is determined based on the weight coefficients, and the similarity between the communication data features and the features of each communication package is calculated using the preset correlation coefficient algorithm. The algorithm for determining the preset correlation coefficient based on the weighting coefficient includes: The Pearson correlation coefficient algorithm is improved based on the weighting coefficients, and the improved Pearson correlation coefficient algorithm is determined as the preset correlation coefficient algorithm.

2. The method according to claim 1, characterized in that, The step of determining the target communication package to recommend to the user based on the selection probability and the similarity includes: For any of the communication packages, the selection probability and similarity corresponding to the communication package are weighted and averaged to obtain the weighted average value corresponding to the communication package. The maximum value is determined from the weighted average values ​​corresponding to each of the aforementioned communication packages, and the communication package corresponding to the maximum value is determined as the target communication package.

3. The method according to claim 1, characterized in that, The pre-trained neural network model is trained in the following way: Identify the sample users to be used as training samples, determine the basic information of the sample users, and determine the target communication packages that the sample users have subscribed to. Extract the basic characteristics of the sample users from the basic information of the sample users; The model is trained using the basic features of the sample users and the target communication package of the sample to obtain the model weight parameters of the neural network model; the model weight parameters include a matrix vector and an offset vector.

4. The method according to claim 3, characterized in that, The determination of sample users as training samples includes: Acquire candidate sample users, determine the candidate sample communication data information of the candidate sample users, and determine the candidate sample target communication package information of the candidate sample target communication packages that the candidate sample users have subscribed to; For any candidate sample user, a correlation analysis is performed on the candidate sample communication data information and the candidate sample target communication package information corresponding to the candidate sample user to obtain the communication data correlation value corresponding to the candidate sample user. The communication data correlation value corresponding to each of the candidate sample users is determined, and the candidate sample users whose communication data correlation value is greater than a preset correlation threshold are determined as the sample users as training samples.

5. The method according to claim 1, characterized in that, The process of determining the user's basic characteristics for the user's basic information, determining the communication data characteristics for the communication data information, and determining the communication package characteristics for the communication package information includes: The user's basic information is subjected to word embedding processing to obtain the corresponding user basic features; The communication data information is subjected to deviation standardization processing to obtain the corresponding communication data features; The communication package information is subjected to deviation standardization processing to obtain the corresponding communication package features.

6. A recommendation device, characterized in that, The device includes: The acquisition module is used to acquire communication package information corresponding to each communication package, as well as basic user information and communication data information of the user to be recommended; The first determining module is used to determine the user basic characteristics for the user basic information, the communication data characteristics for the communication data information, and the communication package characteristics for the communication package information respectively. The input / output module is used to input the user's basic features into a pre-trained neural network model and output the probability of the user to be recommended choosing each of the communication packages. The calculation module is used to calculate the similarity between the communication data features and the features of each of the communication packages using a preset correlation coefficient algorithm; The second determining module is used to determine the target communication package to be recommended to the user to be recommended based on the selection probability and the similarity. The communication data features include voice data features, traffic data features, SMS data features, and value-added service data features. The calculation module includes: The third determining submodule is used to determine the target age group to which the age of the user to be recommended belongs, and to determine the weight coefficients to be assigned to the voice data features, the traffic data features, the SMS data features and the value-added service data features in the target age group respectively; The determination and calculation submodule is used to determine the preset correlation coefficient algorithm based on the weight coefficient, and to calculate the similarity between the communication data features and the features of each communication package using the preset correlation coefficient algorithm; The third determining submodule includes: The Pearson correlation coefficient algorithm is improved based on the weighting coefficients, and the improved Pearson correlation coefficient algorithm is determined as the preset correlation coefficient algorithm.

7. An electronic device, characterized in that, include: A processor, a memory, and a computer program stored in the memory and capable of running on the processor, wherein the computer program, when executed by the processor, implements the steps of a recommended method as described in any one of claims 1-5.

8. 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 a recommended method as described in any one of claims 1-5.

Images