Method, device, apparatus and computer-readable medium for matching data

Abstract

The present invention discloses a method, device, equipment and computer-readable medium for matching data, and relates to the field of computer technology. A specific implementation of the method includes: for a user who has acquired data, judging whether the user will acquire data again according to user information, parameters of acquired data and browsing history of the user, combined with a binary classification model; using a preset classification model, dividing the user into multiple categories according to the probability of acquiring data again; for classified users whose probability is greater than a preset threshold, determining the data to be acquired for the classified user and similar users of the classified user in combination with the historical browsing history and historical acquisition history of the classified user; matching data for the classified user according to the data to be acquired for the classified user, the similar users, the historical browsing time and historical browsing channels of the classified user. This implementation can improve the accuracy of matching data.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a method, device, equipment and computer-readable medium for matching data. Background Art

[0002] With the advent of the wave of artificial intelligence technology, the application of machine learning and deep learning has become increasingly mature, and industries combined with artificial intelligence have gradually developed into a new form protected by intelligence, digitalization, and technology.

[0003] In the process of implementing the present invention, the inventors found that there are at least the following problems in the prior art: using a conventional data matching method to randomly select customers and send messages uniformly, resulting in a low accuracy rate of matching data. Summary of the invention

[0004] In view of this, embodiments of the present invention provide a method, apparatus, device and computer-readable medium for matching data, which can improve the accuracy of matching data.

[0005] To achieve the above object, according to one aspect of an embodiment of the present invention, a method for matching data is provided, comprising:

[0006] For users who have acquired data, judging whether the user will acquire data again is performed based on the user information, the parameters of the acquired data and the browsing history of the user, combined with the binary classification model;

[0007] Using a preset classification model, the users are divided into multiple categories according to the probability of obtaining data again;

[0008] For the classified users whose probability is greater than a preset threshold, combining the historical browsing records and historical acquisition records of the classified users, determining the data to be acquired for the classified users and the similar users of the classified users;

[0009] According to the to-be-acquired data of the classified user, the similar users, the historical browsing time and the historical browsing channel of the classified user, data is matched for the classified user.

[0010] The user information includes one or more of the following: user identification, age, gender and income weight;

[0011] The user browsing record includes one or more of the following: the total number of browsing times within a preset time period, the total browsing time within a preset time period, the number of visited pages within a preset time period, and whether the user is active for N days, where N is a real number greater than or equal to 1.

[0012] The preset classification model is determined based on the level of re-acquired training data, user information, user browsing history and acquired training data, and the level of re-acquired training data is determined based on the repurchase rate or the number of purchased policies.

[0013] The similar users of the classified users are determined based on the historical browsing records and historical acquisition records of the classified users whose probabilities are greater than a preset threshold.

[0014] The step of determining the to-be-acquired data of the classified user and similar users of the classified user in combination with the historical browsing records and historical acquisition records of the classified user for the classified user with the probability greater than the preset threshold comprises:

[0015] For the classified users whose probability is greater than a preset threshold, the data to be acquired for the classified users is determined by combining the historical browsing records, the scores of browsing behaviors, the historical acquisition records and the scores of acquisition behaviors of the classified users;

[0016] Based on the to-be-acquired data of the classified user, similar users to the classified user are selected.

[0017] According to the to-be-acquired data of the classified user, the similar users, the historical browsing time and the historical browsing channel of the classified user, matching data for the classified user includes:

[0018] Determine the preset time and preset channel for sending data according to the historical browsing time and historical browsing channels of the classified users;

[0019] The to-be-acquired data of the classified user and the acquired data of the similar user are used as matching data;

[0020] The matching data is sent according to the preset channel and the preset time.

[0021] Before the binary classification model is combined to determine whether the customer user will obtain data again, the method further includes:

[0022] Among the user information, the parameters of the acquired data and the user browsing history, parameters with high correlation are eliminated.

[0023] According to a second aspect of an embodiment of the present invention, there is provided a device for matching data, including:

[0024] A judgment module, for judging whether a user who has acquired data will acquire data again according to user information, parameters of acquired data and browsing history of the user, in combination with a binary classification model;

[0025] A classification module, used to classify the users into multiple categories according to the probability of obtaining data again using a preset classification model;

[0026] A data module is used to determine the data to be obtained of the classified user and similar users of the classified user in combination with the historical browsing records and historical acquisition records of the classified user for the classified user whose probability is greater than a preset threshold;

[0027] The matching module is used to match data for the classified users according to the to-be-acquired data of the classified users, the similar users, the historical browsing time and the historical browsing channels of the classified users.

[0028] According to a third aspect of an embodiment of the present invention, there is provided an electronic device for matching data, including:

[0029] one or more processors;

[0030] a storage device for storing one or more programs,

[0031] When the one or more programs are executed by the one or more processors, the one or more processors implement the above method.

[0032] According to a fourth aspect of an embodiment of the present invention, there is provided a computer-readable medium on which a computer program is stored, and when the program is executed by a processor, the method as described above is implemented.

[0033] An embodiment of the above invention has the following advantages or beneficial effects: for users who have acquired data, according to user information, parameters of acquired data and browsing history of the user, combined with a binary classification model, it is judged whether the user will acquire data again; using a preset classification model, the users are divided into multiple categories according to the probability of acquiring data again; for classified users whose probability is greater than a preset threshold, the data to be acquired for the classified user and similar users of the classified user are determined in combination with the historical browsing records and historical acquisition records of the classified user; according to the data to be acquired for the classified user, the similar users, the historical browsing time and historical browsing channels of the classified user, data is matched for the classified user.

[0034] First, the binary classification model is used to filter out users who will obtain data again from the acquired data. After excluding a large number of users who do not obtain data again, the preset classification model is used to classify users who obtain data again. The number of users classified above is much smaller than the number of original users, thus shortening the time for matching data.

[0035] Secondly, for users who obtain data again, on the one hand, the data to be obtained is predicted based on the user himself, and on the other hand, data is obtained from similar users, thereby improving the accuracy of matching data.

[0036] The further effects of the above-mentioned non-conventional optional manner will be described below in conjunction with the specific implementation manner. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] The accompanying drawings are used to better understand the present invention and do not constitute an improper limitation of the present invention.

[0038] Figure 1 is a schematic diagram of the main process of the method for matching data according to an embodiment of the present invention;

[0039] Figure 2 is a schematic diagram of a process for determining whether a user will obtain data again according to an embodiment of the present invention;

[0040] Figure 3 is a schematic diagram of a process for determining to-be-acquired data of a classified user and similar users of the classified user according to an embodiment of the present invention;

[0041] Figure 4 is a schematic diagram of a process for classifying user matching data according to an embodiment of the present invention;

[0042] Figure 5 is a schematic diagram of screening users according to an embodiment of the present invention;

[0043] Figure 6 is a schematic diagram of a sending task according to an embodiment of the present invention;

[0044] Figure 7 is a schematic diagram of the main structure of a device for matching data according to an embodiment of the present invention;

[0045] Figure 8 is an exemplary system architecture diagram to which embodiments of the present invention may be applied;

[0046] Fig. 9 It is a schematic diagram of the structure of a computer system of a terminal device or a server suitable for implementing an embodiment of the present invention. DETAILED DESCRIPTION

[0047] The following is a description of exemplary embodiments of the present invention in conjunction with the accompanying drawings, including various details of the embodiments of the present invention to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present invention. Similarly, for clarity and conciseness, the description of well-known functions and structures is omitted in the following description.

[0048] Artificial intelligence has been gradually applied to insurance service processes and all aspects of insurance companies' operations, such as: personalized insurance products for each individual, user-based dynamic product pricing, risk control to reduce claims risk, OCR recognition for fast auto insurance claims, anti-fraud models to reduce payment risks, and NLP to standardize agent terminology to gain insight into telemarketing user preferences.

[0049] The intersection of customized products for different people and multi-format precision reach is an important area of ​​application of artificial intelligence in the insurance industry. With the help of artificial intelligence, users are grouped, the preferences of different groups of users are understood, and future trends are predicted. We are committed to precise multi-format marketing and improving user recall rates.

[0050] At present, the conventional way to match data is to randomly select a group of invalid users and send messages to them uniformly to achieve data matching, thereby achieving marketing for users. The above blind marketing method has poor marketing effect and high operating costs because it does not tap into the real needs of users.

[0051] In order to solve the problem of low accuracy of matching data, the following technical solution in the embodiment of the present invention can be adopted.

[0052] See also Figure 1 , Figure 1 is a schematic diagram of the main process of the method for matching data according to an embodiment of the present invention, by analyzing the user whose data has been acquired, data matching is achieved from two aspects: the user himself and similar users. Figure 1 As shown, the specific steps include:

[0053] S101. For a user who has acquired data, judging whether the user will acquire data again is performed based on the user information, parameters of the acquired data, and the user's browsing history in combination with a binary classification model.

[0054] In an embodiment of the present invention, data is matched again for users who have acquired data. The users who have acquired data are users who already have records of acquiring data. As an example, users who have acquired data include users who have purchase records, that is, the number of purchase records is greater than or equal to 1.

[0055] See also Figure 2 , Figure 2 2 is a flow chart of determining whether a user will obtain data again according to an embodiment of the present invention, which specifically includes the following steps:

[0056] S201. Clean the original users whose data have been acquired to obtain the users whose data have been acquired.

[0057] The users whose data has been acquired involve a lot of data, but not every user's user information is complete. Therefore, the original users whose data has been acquired need to be cleaned to obtain the users whose data has been acquired. It can be understood that the users whose data has been acquired are the users obtained after cleaning and elimination based on the original users whose data has been acquired.

[0058] The following is a brief introduction to the specific process of cleaning raw data.

[0059] The features of the original users of the acquired data are stored on the Hadoop platform. As an example, the features of the original users of the acquired data include one or more of the following: user ID, policyholder age, policyholder gender, income weight, number of policies, total number of views, total viewing time, number of visited pages, whether active for 30 days, whether active for 60 days, etc.

[0060] If the characteristics of the original users of the acquired data are incomplete, that is, there are missing values, the missing values ​​need to be processed. As an example, one or more of the following methods are used to process the missing values: zero filling, zero value difference, mean interpolation, median interpolation, fixed value interpolation, adjacent value interpolation, and Lagrange interpolation.

[0061] After processing the missing values, the features of the original users can be cleaned according to the preset parameter interval. For the preset parameter interval of the features of the original users, the original users whose data have been obtained are cleaned to obtain the original users within the preset parameter interval, that is, the users whose data have been obtained.

[0062] S202: For users who have acquired data, judging whether the user will acquire data again is performed based on the user information, parameters of the acquired data, and the user's browsing history in combination with a binary classification model.

[0063] In an embodiment of the present invention, a binary classification model is used to determine whether the user will obtain data again. The binary classification model used is a trained model. As an example, the binary classification model is trained on the basis of the following algorithms: Adaboost algorithm, GBDT algorithm, SVM algorithm, LR algorithm, RF algorithm and Xgboost algorithm. In the process of training the model, AUC, precision rate and recall rate are used to measure the training effect of the model.

[0064] For the original user who has acquired the data, the user information, the parameters of the acquired data and the user's browsing history are input into the binary classification model to determine whether the user will acquire the data again.

[0065] Among them, user information includes user ID, insured person's age, insured person's gender and income weight. Income weight includes the proportion of insured person's income to family income. The parameters of the acquired data include: the number of purchased policies and the purchase price of purchased policies. User browsing history includes: the total number of browsing times within a preset time period, the total browsing time within a preset time period, the number of visited pages within a preset time period, whether N days are active and whether N days are active, where N is a real number greater than or equal to 1. The value of N can be preset according to the actual application scenario.

[0066] exist Figure 2 In the embodiment, for the original users after cleaning, a binary classification model is combined to determine the users who will obtain data again.

[0067] In the embodiment of the present invention, a binary classification model is used to classify users into two categories, the first category is users who do not obtain data again, and the second category is users who will obtain data again. For the first category of users, there is no need to match data; for the second category of users, it may be necessary to match data. It can be seen that the role of the binary classification model is to eliminate users who do not need to match data, and then filter out users who need to match data.

[0068] S102: Using a preset classification model, the users are divided into multiple categories according to the probability of obtaining data again.

[0069] In an embodiment of the present invention, a preset classification model is used to classify users into multiple categories according to the probability of re-acquiring data. It should be noted that the preset classification model is a trained model. As an example, the preset classification model is established based on the XGBoost algorithm or the RF algorithm.

[0070] In one embodiment of the present invention, the preset classification model is determined based on the level of re-acquired training data, user basic information, user browsing information and acquired training data.

[0071] In the embodiment of the present invention, a preset classification model is used to classify users into multiple categories according to the probability of re-acquiring data. It should be noted that the classified users are not the original users, but the users who will re-acquire data. In other words, compared with the original customers, the number of users who will re-acquire data is greatly reduced, so the classification using the preset classification model can improve accuracy and shorten classification time.

[0072] The following is an exemplary description of the process of establishing a preset classification model.

[0073] The level of re-acquiring training data is the level at which the user re-acquires training data. The level of re-acquiring training data may be determined based on the repurchase rate or the number of policies purchased.

[0074] As an example, the levels of training data acquired again include level 1, level 2, and level 3.

[0075] Level 1: Repurchase rate ≤ 60%; or, the number of policies is 2 or 3.

[0076] Level 2: 60% < repurchase rate ≤ 80%; or, 3 policies < number of policies < 12 policies.

[0077] Level 3: Repurchase rate ≥ 80%; or, number of policies ≥ 12.

[0078] The preset classification model is determined based on the level of reacquired training data, user information, user browsing history and acquired training data. That is, the preset classification model is trained using the level of reacquired training data and acquired training data involved in the training data.

[0079] After obtaining the preset classification model, the following are input into the preset classification model: the level of re-acquiring data, user information, user browsing history, and acquired training data. The preset classification model outputs the probability of the user re-acquiring data. Then, according to the probability of re-acquiring data, the users are divided into multiple categories.

[0080] As an example, based on a preset threshold, users are divided into a category with a probability greater than the preset threshold and a category with a probability less than or equal to the preset threshold. The purpose of this is that users whose probability of obtaining data again is greater than the preset threshold tend to obtain data again. By selecting users by using the preset threshold, the time spent on matching data is shortened while reducing the number of users matching data.

[0081] In one embodiment of the present invention, taking the example that the parameters of the acquired data include the policyholder purchasing an insurance policy, an exemplary description is given by comparing the XGBoost algorithm or the RF algorithm.

[0082] The insured who purchases two or more policies is defined as a "repeat purchaser". Among them, renewal customers are also considered repeat purchasers, and the average premium of the policy is more than 80 yuan. The formula for the repeat purchase rate is as follows:

[0083] n-time repurchase rate = number of people who repurchased for the nth time / (number of people who repurchased for the n-1th time), n = 2, 3, 4, ...

[0084] The classification of repurchase rates is shown in Table 1, which is a schematic diagram of the classification of repurchase rates for insurance policies as samples:

[0085] Table 1

[0086]

[0087]

[0088] 129,144 policy samples were extracted, with level 1: level 2: level 3 = 4:3:3. There are 57,276 samples at level 1, 35,000 samples at level 2, and 36,868 samples at level 3. See Table 2 for details:

[0089] Table 2

[0090] Repurchase level Number of people Proportion 1 57276 44.4％ 2 35000 27.1％ 3 36868 28.5％ total 129144 100.0％

[0091] Conduct data statistics and analysis on the repurchase probability of the above samples, calculate the coverage of the samples, pre-process the data using the zero-filling method, and analyze the correlation between various indicators.

[0092] With the help of XGBoost algorithm and RF algorithm training, after comparative analysis of various indicators, it is found that the effect of RF algorithm is better than that of XGBoost algorithm. The specific comparative effect is shown in Table 3. Table 3 is the comparison of indicators for evaluating XGBoost algorithm and RF algorithm.

[0093] Table 3

[0094] Evaluation method XGBoost RF MSE 0.9721 0.9338 Kappa value 0.9540 0.96900

[0095] Mse: represents the mean square error. The smaller the error, the better. This shows that the RF algorithm is better than the GBDT model algorithm.

[0096] The Kappa coefficient is an indicator to measure classification accuracy, and the calculation formula is: K = (P0-Pe) / (1-Pe); P0 is the actual consistency rate, and Pe is the theoretical consistency rate. It can be seen that the RF algorithm is better than the XGBoost algorithm.

[0097] In the above embodiment, the binary classification model is used to determine whether the user will obtain data again; the preset classification model is used to determine the probability of the user obtaining data again; and the preset threshold is used to select users who need to match data.

[0098] Since the training data of the binary classification model and the preset classification model are different, the functions of the binary classification model and the preset classification model are different. First, the binary classification model is used to filter out users who will obtain data again; then the preset classification model is used to determine the probability of users obtaining data again. Finally, the preset threshold is used to select users who need to match data.

[0099] Using binary classification models, preset classification models and preset thresholds for layer-by-layer screening can reduce the number of users while improving the speed and accuracy of user classification.

[0100] S103 . For a classified user whose probability is greater than a preset threshold, determine the data to be obtained of the classified user and similar users of the classified user in combination with the historical browsing records and historical acquisition records of the classified user.

[0101] The classified users with a probability greater than the preset threshold have a greater possibility of acquiring data again. Then, the data to be acquired and similar users are determined from two aspects, namely, historical browsing records and historical acquisition records. That is, the data to be acquired is determined based on the historical browsing records and historical acquisition records of the classified users with a probability greater than the preset threshold; similar users are determined based on the historical browsing records and historical acquisition records of the classified users with a probability greater than the preset threshold. As an example, the preset threshold is 50%.

[0102] It should be noted that the technical solution involved below is implemented on the basis of classified users whose probability of obtaining data again is greater than a preset threshold. Therefore, the classified users involved below refer to: classified users whose probability of obtaining data again is greater than a preset threshold.

[0103] See also Figure 3 , Figure 3 1 is a flow chart of determining the to-be-acquired data of a classified user and similar users of the classified user according to an embodiment of the present invention, which specifically includes the following steps:

[0104] S301 . For a classified user whose probability is greater than a preset threshold, determine the data to be acquired of the classified user in combination with the historical browsing records and historical acquisition records of the classified user.

[0105] The data to be acquired by the classified user is the data that the user is more likely to acquire. From the perspective of the classified user himself, the possibility of acquiring the data that has been browsed and / or acquired is greater.

[0106] Considering that this is not the first time for the classified user to obtain data, it is highly likely that there are historical browsing records and historical acquisition records. Therefore, the data to be obtained for the classified user can be determined by combining the historical browsing records and historical acquisition records of the classified user.

[0107] Of course, for classified users whose historical browsing records and / or historical acquisition records are empty, historical browsing records and / or historical acquisition records can be preset to determine the data to be acquired. As an example, the preset historical browsing records include browsing records of the user's frequently used commodities. The historical acquisition records include purchase records of the user's frequently used commodities.

[0108] In the embodiment of the present invention, the historical browsing records of the classified users specifically include the number of historical browsing times and the historical browsing duration. The historical browsing duration refers to the duration of a browsing process. As an example, the number of historical browsing times includes two times, the historical browsing duration of the first historical browsing time is 10 minutes; the historical browsing duration of the second historical browsing time is 5 minutes.

[0109] The score of the browsing behavior is determined based on the number of historical browsing times and the historical browsing time. As an example, the score of the browsing behavior = the number of historical browsing times × 0.3 + the historical browsing time × 0.7. Among them, 0.3 and 0.7 are preset weights, and the preset weights can be adjusted according to actual needs.

[0110] In an embodiment of the present invention, the historical acquisition record includes the number of items and the unit price. Based on the number of items and the unit price, the score of the acquisition behavior is determined. As an example, the score of the acquisition behavior = the number of products × 0.5 + the unit price × 0.5.

[0111] Combine the above-mentioned browsing behavior score and acquisition behavior score to finally determine the product score. Among them, the browsing behavior weight and the acquisition behavior weight are preset. As an example, the browsing behavior weight = 0.4, and the acquisition behavior weight = 0.6. According to the product score, the data of the first K products are used as the data to be acquired for the classified users. K is a preset parameter greater than 0.

[0112] It is understandable that the data to be acquired for the classified users is highly correlated with the products that have been browsed and acquired.

[0113] S302: Select similar users to the classified users based on the to-be-acquired data of the classified users.

[0114] In the embodiment of the present invention, on the one hand, starting from the historical records of the classified users, the data to be acquired is determined; on the other hand, starting from the users, similar users to the classified users are determined.

[0115] Specifically, the data to be obtained of the classified users is used as the basis for determining similar users. Generally speaking, if two users obtain the same data, it is highly likely that the two users are similar users.

[0116] As an example, for the preference vector of all category data of user A and the preference vector of all category data of user B, the angle in multidimensional space can be calculated using the vector cosine formula. Among them, the preference vector is a vector composed of multiple categories of data. Each element in the vector represents the user's acquisition of the data. For example: 1 represents acquisition; 0 represents non-acquisition. Then, the cosine similarity is used to measure the similarity between user A and user B. As an example, the data of products that the user has purchased is marked as acquired; the data of products that the user has not purchased is marked as non-acquisition.

[0117] The following is a brief introduction to cosine similarity.

[0118] The principle of cosine similarity is to use the cosine function. A vector is a line segment with an arrow in space. The smaller the angle between two vectors, the closer the similarity of the two vectors. The vector of multiple attributes is represented by r(x1,x2,...,xn ).

[0119] For user A's preference vector r for all categories of products A (A1,A2,...,A n ) and user B’s vector r B (B1,B2,...,B n ), the angle between the two in multidimensional space can be calculated using the vector cosine formula:

[0120]

[0121] The cosine value range is between -1 and 1, where 0 means completely vertical 90°, -1 means the angle is 180°, and 1 means the angle is 0°. The closer the coefficient is to 1, the smaller the vector angle is, and the higher the correlation between the two users is.

[0122] For user A and user B, their similarity is:

[0123]

[0124] As an example, the correspondence between the similarity degree and the cosine similarity includes:

[0125] Very similar: 0.8-1.0;

[0126] Similar: 0.6-0.8;

[0127] Normal: 0.4-0.6;

[0128] Not very similar: 0-0.4;

[0129] Completely dissimilar: -1.0—0.

[0130] Based on the data to be obtained for the classified user, similar users of the classified user are obtained and determined according to the cosine similarity. As an example, users with cosine similarity between 0.8 and 1.0 are regarded as similar users of the classified user.

[0131] It is understandable that the cosine similarity between the classified user and other users is determined based on the data to be obtained of the classified user. Based on the above cosine similarity and the corresponding relationship between the similarity degree and the cosine similarity, similar users of the classified user are selected. Considering that the data to be obtained of the classified user is determined by the historical browsing records and the historical acquisition records. Then, similar users are determined based on the historical browsing records and the historical acquisition records of the classified users whose probability is greater than a preset threshold.

[0132] exist Figure 3In an embodiment, for classified users whose probability is greater than a preset threshold, data to be acquired and similar users are determined based on historical browsing records and historical acquisition records to improve the accuracy and richness of matching data.

[0133] S104: Match data for the classified user according to the to-be-acquired data of the classified user, similar users, historical browsing time and historical browsing channels of the classified user.

[0134] Matching data for classified users involves data, time and channels, that is, matching data for classified users is achieved based on data at a preset time and in a preset channel.

[0135] See also Figure 4 , Figure 4 1 is a flow chart of classifying user matching data according to an embodiment of the present invention, which specifically includes the following steps:

[0136] S401. Determine a preset time and preset channel for sending data according to the historical browsing time and historical browsing channels of the classified users.

[0137] For classified users, they usually browse data at fixed times. The preset time is determined according to the historical browsing time of the classified users. As an example, classified user A browses data at 8 am and 8 pm, so the preset time for sending data is determined to be 8 am and 8 pm.

[0138] Users can obtain data through a variety of browsing channels. As an example, the channels include one or more of the following: active acquisition by application (APP), APP push, active acquisition by web page, web page push, and telephone communication.

[0139] The preset channel is determined according to the historical browsing channels of the classified users. As an example, classified user A usually receives data pushed by APP; classified user B answers the phone to obtain data.

[0140] In one embodiment of the present invention, it is considered that classified users may obtain data through different browsing channels at different browsing times. As an example, user A obtains data through APP push at 8 am, user A actively obtains data through the web page at 4 pm, and user A actively obtains data from the APP at 9 pm.

[0141] In view of the above situation, the preset time and preset channel for sending data can be determined according to the historical browsing time and the historical browsing channel. That is to say, if the preset time for sending data is different, the preset channel may be different.

[0142] S402: Use the to-be-acquired data of the classified user and the acquired data of similar users as matching data.

[0143] The data obtained by similar users of the classified user is not completely the same as the data obtained by the classified user, so the data obtained by the similar users that is different from the data obtained by the classified user can be used as matching data.

[0144] In addition, the data to be acquired by the classified user is data that the classified user is more likely to acquire. Furthermore, the data to be acquired by the classified user can also be used as matching data.

[0145] It can be understood that the matching data consists of two parts. One part is the data that has been acquired by similar users and is not the data that has been acquired by the classified users; the other part is the data to be acquired by the classified users.

[0146] S403: Send matching data according to a preset channel and preset time.

[0147] After determining the preset time and the preset channel for sending the data, the matching data can be sent. As an example, in order to remind the user, the same matching data can be sent multiple times. For example: at the first preset time, through the first preset channel, matching data 1 is sent; at the second preset time, through the second preset channel, matching data 2 is sent.

[0148] exist Figure 4 In the embodiment, matching data is assigned to the user through a preset time and a preset channel. Since the time and channel for sending matching data are consistent with user habits, the matching accuracy can be improved.

[0149] The following takes user matching data as an example to illustrate how to send matching data according to a preset channel and preset time.

[0150] First, the classified users whose probability of re-acquiring data is greater than a preset threshold are screened out. The data to be acquired is determined based on the ID, historical browsing records, and historical acquisition records of the above classified users. As an example, the data is specifically a product. Based on the similarity between the products involved in the user's historical browsing records and historical acquisition records and other products, a collaborative filtering algorithm can be used to calculate a list of products that the user likes. In addition, products that are different from the products that the user has acquired among the products that similar users have acquired can be added to the product list. In other words, the product list includes matching data.

[0151] See also Figure 5 , Figure 5 is a schematic diagram of screening users according to an embodiment of the present invention. Figure 5 This involves filtering users. Figure 5 The software interface shown is implemented. It should be noted that Figure 5 Screened users are classified users whose probability is greater than a preset threshold, that is, users who need matching data.

[0152] Figure 5 You need to fill in or select multiple items. For example: customer group name, customer group description and screening conditions. Among them, the screening conditions can be selected from the optional indicators. The optional indicators include: total number of customers, ID cards, reachable number, number of mobile phones, number of devices and number of mailboxes. Among them, the number of devices includes not only the number of mobile phones, but also the number of other terminals, such as tablets and personal computers. The products in the above product list are sold to users corresponding to the ID after configuring the tags on the sales platform.

[0153] As for the way to match data, that is, the way to sell products, the following strategies can be adopted:

[0154] Strategy 1: Develop the recommended time for each user based on purchase hour preference.

[0155] Strategy 2: Develop a recommendation channel for each user based on the platform preference for purchasing products.

[0156] Strategy three: Set the time for each user to send messages based on the browsing time preferences of browsing users.

[0157] The channels for delivering data to users include: one or more of: active acquisition through APP, push through APP, active acquisition through web pages, push through web pages, and telephone communication.

[0158] See Table 4, which is a statistical table of sales channels corresponding to products. Among them, the sales plan has been marked and distinguished, APP1 is application 1; PC is data sent through personal computers; WAP is data sent through mobile terminal browsers; health APP is health APP; official website APP is the official website of the sales product; social APP is social APP.

[0159] Table 4

[0160]

[0161]

[0162] For each product, the sales plan in Table 4 can be combined to select the first few products in the product list, send them through channels according to the recommendation time of each customer, and provide personalized sales to users in a timely and targeted manner.

[0163] In the sales plan in Table 4, it can be seen from statistics that sales plan S20180087 is better because it has the highest sales volume in various channels. Therefore, sales plan S20180087 can be used to sell products.

[0164] As an example, ABTest can be used to monitor the effect. For example, by selecting different recommendation times, ABTest can be implemented to determine the better recommendation time; another example is by selecting different recommendation channels, ABTest can be implemented to determine the better recommendation channel.

[0165] See also Figure 6 , Figure 6 1 is a schematic diagram of sending a task according to an embodiment of the present invention. Data is matched for a user in the manner of sending a task.

[0166] Figure 6 You need to fill in the following options. Task type, task name, delivery mode, customer group name, delivery channel and message template. Task types include single task and task group. Delivery mode can be selected as immediate delivery or delivery at a preset time. Customer group name is Figure 5 The content filled in. The sending channel can be selected from APP, web page and phone. The message template refers to the template corresponding to the message sent. As an example, the message model can select APP and notification content.

[0167] In one embodiment of the present invention, by including Figure 5 and Figure 6 The intelligent marketing platform sends matching data to users according to the preset channels and preset time. In the intelligent marketing platform, you can fill in the task type, preset channel and preset time. For example, the task type includes single task and task group.

[0168] In one embodiment of the present invention, the binary classification model screens out users who will obtain data again; the preset classification model determines the probability of users obtaining data again. After matching data for the classified users, the matching data can be used as successfully matched data to update the binary classification model and the preset classification model.

[0169] As an example, a binary classification model and a preset classification model are updated in a preset period. The update period of the binary classification model may be different from the update period of the preset classification model. As another example, after determining the successfully matched data, the binary classification model and / or the preset classification model may be updated in real time with the successfully matched data.

[0170] In the above embodiment, for users who have acquired data, according to the user information, the parameters of the acquired data and the user's browsing history, combined with a binary classification model, it is judged whether the user will acquire data again; using a preset classification model, the users are divided into multiple categories according to the probability of acquiring data again; for classified users whose probability is greater than a preset threshold, the data to be acquired for the classified user and the similar users of the classified user are determined in combination with the historical browsing records and historical acquisition records of the classified user; according to the data to be acquired for the classified user, the similar users, the historical browsing time and historical browsing channels of the classified user, data is matched for the classified user.

[0171] First, the binary classification model is used to filter out users who will obtain data again from the acquired data. After excluding a large number of users who do not obtain data again, the preset classification model is used to classify users who obtain data again. The number of users classified above is much smaller than the number of original users, thus shortening the time for matching data.

[0172] Secondly, for users who obtain data again, on the one hand, the data to be obtained is predicted based on the user himself, and on the other hand, data is obtained from similar users, thereby improving the accuracy of matching data.

[0173] See also Figure 7 , Figure 7 is a schematic diagram of the main structure of the device for matching data according to an embodiment of the present invention. The device for matching data can implement a method for matching data, such as Figure 7 As shown, the device for matching data specifically includes:

[0174] The judgment module 701 is used to judge whether the user who has acquired data will acquire data again according to the user information, the parameters of the acquired data and the browsing history of the user in combination with the binary classification model;

[0175] A classification module 702, configured to use a preset classification model to classify the user into multiple categories according to the probability of obtaining data again;

[0176] Data module 703, for determining the data to be obtained of the classified user and similar users of the classified user in combination with the historical browsing records and historical acquisition records of the classified user for the classified user with the probability greater than a preset threshold;

[0177] The matching module 704 is used to match data for the classified user according to the to-be-acquired data of the classified user, the similar users, the historical browsing time and the historical browsing channel of the classified user.

[0178] In one embodiment of the present invention, the user information includes one or more of the following: user identification, age, gender and income weight;

[0179] The user browsing record includes one or more of the following: the total number of browsing times within a preset time period, the total browsing time within a preset time period, the number of visited pages within a preset time period, and whether the user is active for N days, where N is a real number greater than or equal to 1.

[0180] In one embodiment of the present invention, the preset classification model is determined based on the level of re-acquired training data, user information, user browsing history and acquired training data, and the level of re-acquired training data is determined based on the repurchase rate or the number of purchased policies.

[0181] In one embodiment of the present invention, similar users of the classified user are determined based on historical browsing records and historical acquisition records of the classified user with a probability greater than a preset threshold.

[0182] In one embodiment of the present invention, the data module 703 is specifically used to determine the data to be acquired for the classified user whose probability is greater than a preset threshold, combining the historical browsing records, browsing behavior scores, historical acquisition records and acquisition behavior scores of the classified user;

[0183] Based on the to-be-acquired data of the classified user, similar users to the classified user are selected.

[0184] In one embodiment of the present invention, the matching module 704 is specifically configured to determine the preset time for sending data and the preset channel for the preset time according to the historical browsing time and historical browsing channel of the classified user;

[0185] The to-be-acquired data of the classified user and the acquired data of the similar user are used as matching data;

[0186] The matching data is sent according to the preset channel and the preset time.

[0187] In one embodiment of the present invention, the judgment module 701 is further configured to remove highly correlated parameters from the user information, the parameters of the acquired data, and the user browsing history.

[0188] Figure 8 An exemplary system architecture 800 is shown to which the method for matching data or the apparatus for matching data according to the embodiments of the present invention may be applied.

[0189] like Figure 8As shown, system architecture 800 may include terminal devices 801, 802, 803, a network 804, and a server 805. Network 804 is used to provide a medium for communication links between terminal devices 801, 802, 803 and server 805. Network 804 may include various connection types, such as wired, wireless communication links, or optical fiber cables, etc.

[0190] Users can use terminal devices 801, 802, and 803 to interact with server 805 through network 804 to receive or send messages, etc. Various communication client applications can be installed on terminal devices 801, 802, and 803, such as shopping applications, web browser applications, search applications, instant messaging tools, email clients, social platform software, etc. (only examples).

[0191] The terminal devices 801 , 802 , and 803 may be various electronic devices having a display screen and supporting web browsing, including but not limited to smart phones, tablet computers, laptop computers, desktop computers, and the like.

[0192] The server 805 may be a server that provides various services, such as a backend management server (only an example) that provides support for shopping websites browsed by users using the terminal devices 801, 802, and 803. The backend management server may analyze and process the received data such as product information query requests, and feed back the processing results (such as target push information, product information - only an example) to the terminal device.

[0193] It should be noted that the data matching method provided in the embodiment of the present invention is generally executed by the server 805 , and accordingly, the data matching device is generally disposed in the server 805 .

[0194] It should be understood that Figure 8 The number of terminal devices, networks and servers in the embodiment is only for illustration. Any number of terminal devices, networks and servers may be provided according to implementation requirements.

[0195] Reference below Fig. 9 , which shows a schematic diagram of the structure of a computer system 900 of a terminal device suitable for implementing an embodiment of the present invention. Fig. 9 The terminal device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention.

[0196] like Fig. 9As shown, the computer system 900 includes a central processing unit (CPU) 901, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 902 or a program loaded from a storage part 908 into a random access memory (RAM) 903. In the RAM 903, various programs and data required for the operation of the system 900 are also stored. The CPU 901, the ROM 902, and the RAM 903 are connected to each other via a bus 904. An input / output (I / O) interface 905 is also connected to the bus 904.

[0197] The following components are connected to the I / O interface 905: an input section 906 including a keyboard, a mouse, etc.; an output section 907 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage section 908 including a hard disk, etc.; and a communication section 909 including a network interface card such as a LAN card, a modem, etc. The communication section 909 performs communication processing via a network such as the Internet. A drive 910 is also connected to the I / O interface 905 as needed. A removable medium 911, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 910 as needed, so that a computer program read therefrom is installed into the storage section 908 as needed.

[0198] In particular, according to the embodiments disclosed in the present invention, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the embodiments disclosed in the present invention include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from the network through the communication part 909, and / or installed from the removable medium 911. When the computer program is executed by the central processing unit (CPU) 901, the above-mentioned functions defined in the system of the present invention are executed.

[0199] It should be noted that the computer-readable medium shown in the present invention may be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present invention, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, device or device. In the present invention, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, which carries a computer-readable program code. This propagated data signal may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Computer-readable signal media may also be any computer-readable medium other than computer-readable storage media, which may send, propagate or transmit a program for use by or in conjunction with an instruction execution system, apparatus or device. The program code contained on the computer-readable medium may be transmitted using any appropriate medium, including but not limited to: wireless, wire, optical cable, RF, etc., or any suitable combination of the above.

[0200] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present invention. In this regard, each box in the flow chart or block diagram can represent a module, a program segment, or a part of a code, and the above-mentioned module, program segment, or a part of a code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flow chart, and the combination of the boxes in the block diagram or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0201] The modules involved in the embodiments of the present invention may be implemented in software or hardware. The modules described may also be set in a processor. For example, they may be described as: a processor includes a judgment module, a classification module, a data module, and a matching module. The names of these modules do not constitute limitations on the modules themselves in certain cases. For example, the judgment module may also be described as "for users who have acquired data, according to user information, parameters of acquired data, and browsing history of the user, combined with a binary classification model, to determine whether the user will acquire data again."

[0202] As another aspect, the present invention further provides a computer-readable medium, which may be included in the device described in the above embodiment; or may exist independently without being assembled into the device. The above computer-readable medium carries one or more programs, and when the above one or more programs are executed by a device, the device includes:

[0203] For users who have acquired data, judging whether the user will acquire data again is performed based on the user information, the parameters of the acquired data and the browsing history of the user, combined with the binary classification model;

[0204] Using a preset classification model, the users are divided into multiple categories according to the probability of obtaining data again;

[0205] For the classified users whose probability is greater than a preset threshold, combining the historical browsing records and historical acquisition records of the classified users, determining the data to be acquired for the classified users and the similar users of the classified users;

[0206] According to the to-be-acquired data of the classified user, the similar users, the historical browsing time and the historical browsing channel of the classified user, data is matched for the classified user.

[0207] According to the technical solution of the embodiment of the present invention, for users who have acquired data, according to user information, parameters of acquired data and browsing history of the user, combined with a binary classification model, it is judged whether the user will acquire data again; using a preset classification model, the users are divided into multiple categories according to the probability of acquiring data again; for classified users whose probability is greater than a preset threshold, the data to be acquired for the classified user and similar users of the classified user are determined in combination with the historical browsing records and historical acquisition records of the classified user; according to the data to be acquired for the classified user, the similar users, the historical browsing time and historical browsing channels of the classified user, data is matched for the classified user.

[0208] First, the binary classification model is used to filter out users who will obtain data again from the acquired data. After excluding a large number of users who do not obtain data again, the preset classification model is used to classify users who obtain data again. The number of users classified above is much smaller than the number of original users, thus shortening the time for matching data.

[0209] Secondly, for users who obtain data again, on the one hand, the data to be obtained is predicted based on the user himself, and on the other hand, data is obtained from similar users, thereby improving the accuracy of matching data.

[0210] The above specific implementations do not constitute a limitation on the protection scope of the present invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions may occur depending on design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for matching data, characterized in that: include: For users who have acquired data, judging whether the user will acquire data again is performed based on the user information, the parameters of the acquired data and the user's browsing history, in combination with the binary classification model; the user who has acquired data is a user who already has a record of acquiring data; Using a preset classification model, the users are divided into multiple categories according to the probability of obtaining data again; The preset classification model is used to determine the probability of a user acquiring data again; For the classified users whose probability is greater than a preset threshold, combining the historical browsing records and historical acquisition records of the classified users, determining the data to be acquired for the classified users and the similar users of the classified users; According to the to-be-acquired data of the classified user, the similar users, the historical browsing time and the historical browsing channel of the classified user, data is matched for the classified user.

2. The method for matching data according to claim 1, characterized in that: The user information includes one or more of the following: user identification, age, gender and income weight; The user browsing record includes one or more of the following: the total number of browsing times within a preset time period, the total browsing time within a preset time period, the number of visited pages within a preset time period, and whether the user is active for N days, where N is a real number greater than or equal to 1.

3. The method for matching data according to claim 1, characterized in that: The preset classification model is determined based on the level of re-acquired training data, user information, user browsing history and acquired training data, and the level of re-acquired training data is determined based on the repurchase rate or the number of purchased policies.

4. The method for matching data according to claim 1, characterized in that: The similar users of the classified users are determined based on the historical browsing records and historical acquisition records of the classified users whose probabilities are greater than a preset threshold.

5. The method for matching data according to claim 1, characterized in that: The step of determining the to-be-acquired data of the classified user and similar users of the classified user in combination with the historical browsing records and historical acquisition records of the classified user for the classified user with the probability greater than the preset threshold comprises: For the classified users whose probability is greater than a preset threshold, the data to be acquired for the classified users is determined by combining the historical browsing records, the scores of browsing behaviors, the historical acquisition records and the scores of acquisition behaviors of the classified users; Based on the to-be-acquired data of the classified user, similar users to the classified user are selected.

6. The method for matching data according to claim 1, characterized in that: According to the to-be-acquired data of the classified user, the similar users, the historical browsing time and the historical browsing channel of the classified user, matching data for the classified user includes: Determine the preset time and preset channel for sending data according to the historical browsing time and historical browsing channels of the classified users; The to-be-acquired data of the classified user and the acquired data of the similar user are used as matching data; The matching data is sent according to the preset channel and the preset time.

7. The method for matching data according to claim 1, characterized in that: Before the binary classification model is combined to determine whether the user will obtain data again, the method further includes: Among the user information, the parameters of the acquired data and the user browsing history, parameters with high correlation are eliminated.

8. A device for matching data, characterized in that: include: A judgment module is used to judge, for a user who has acquired data, whether the user will acquire data again according to the user information, the parameters of the acquired data and the user's browsing history, combined with a binary classification model; the user who has acquired data is a user who already has a record of acquiring data; A classification module, used to classify the users into multiple categories according to the probability of obtaining data again using a preset classification model; The preset classification model is used to determine the probability of a user acquiring data again; A data module is used to determine the data to be obtained of the classified user and similar users of the classified user in combination with the historical browsing records and historical acquisition records of the classified user for the classified user whose probability is greater than a preset threshold; The matching module is used to match data for the classified users according to the to-be-acquired data of the classified users, the similar users, the historical browsing time and the historical browsing channels of the classified users.

9. An electronic device for matching data, characterized in that: include: one or more processors; a storage device for storing one or more programs, When the one or more programs are executed by the one or more processors, the one or more processors implement the method according to any one of claims 1 to 7.

10. A computer readable medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images