Tourism marketing short message sending method and device

Abstract

The invention relates to the technical field of tourism marketing, and discloses a tourism marketing short message sending method and device, and the method comprises the steps: collecting real-time signaling data of a target user located in a target scenic region and other preset type data from an operator; based on the real-time signaling data and the preset type data, key features of the target user are extracted; inputting the key features into a preset LightGBM model for identification, and outputting an identification result of the tourist user; according to the identification result of the tourist user, matching a sending number, a short message content, a marketing area base station and sending time corresponding to the tourist user; and when the tourist user enters the target marketing area, triggering a tourist marketing short message sending task, and sending a tourist marketing short message. The method has the advantages that the pushing precision of the tourism marketing short messages is improved, the problem of short message harassment to local resident people is solved, meanwhile, the short message sending benefit is greatly improved, and the tourism marketing short message expense waste is reduced.

Description

Technical Field

[0001] The present application relates to the field of tourism marketing technology, and in particular to a method and device for sending tourism marketing text messages. Background Art

[0002] In the prior art, roaming welcome SMS messages are often monitored based on base station signaling, and a resident user identification model is constructed based on historical signaling data for filtering, identifying the user group roaming into the base station, and finally obtaining the target number before sending the SMS message through the SMS gateway. However, this method relies on the filtering of resident users, and the judgment of resident users has limitations, resulting in inaccurate target numbers, which affects the push accuracy. Summary of the invention

[0003] In order to improve the push accuracy of tourism marketing SMS, the present application provides a tourism marketing SMS sending method and device.

[0004] In a first aspect, the present application provides a method for sending tourism marketing text messages.

[0005] This application is achieved through the following technical solutions:

[0006] A method for sending tourism marketing SMS includes the following steps:

[0007] Collect real-time signaling data of target users in target scenic spots and other preset data from operators;

[0008] Extracting key features of the target user based on the real-time signaling data and the preset type data;

[0009] Input the key features into the preset LightGBM model for identification, and output the identification results of the tourist users;

[0010] According to the identification result of the tourist user, the sending number, SMS content, marketing area base station and sending time corresponding to the tourist user are matched to complete the pre-configuration of the tourist marketing SMS sending task;

[0011] When a tourist user enters a target marketing area, the tourist marketing SMS sending task is triggered, and a tourist marketing SMS is sent.

[0012] In a preferred example, the present application can be further configured as follows: the steps of constructing the LightGBM model include:

[0013] Initialize the training parameters of the LightGBM model, including the number of trees n_estimators, the maximum depth of the tree max_depth, the learning rate learning_rate, the number of leaf nodes num_leaves of the tree, and the class weight ratio class_weights;

[0014] Input historical key features into the initialized LightGBM model for learning, evaluate the current AUC value and KS value of the model, and optimize the training parameters of the current LightGBM model based on the current AUC value and KS value;

[0015] When the current AUC value and KS value of the model meet the preset conditions, the current LightGBM model is output as the preset LightGBM model.

[0016] In a preferred example, the present application can be further configured as follows: the preset training parameters of the LightGBM model meet the following conditions:

[0017] The number of trees n_estimators = 1000;

[0018] The maximum depth of the tree max_depth = 5;

[0019] Learning rate learning_rate = 0.27;

[0020] The number of leaf nodes in the tree num_leaves = 61;

[0021] Class weight ratio class_weights = [1,300].

[0022] In a preferred example, the present application can be further configured as follows: the real-time signaling data includes VOLTE registration signaling and call signaling, the preset type data includes Internet bill data, user package value data, whether it is integrated package data, whether it is a foreign ID card data and whether the registered residence is rural data, and the real-time signaling data of the target user located in the target scenic spot and the rest of the preset type data from the operator are collected, including:

[0023] Collect VOLTE registration signaling and call signaling of target users in the target scenic area; and

[0024] Collect Internet call bill data from operators, user package value data, whether the package is integrated, whether the ID card data is from another place, and whether the household registration place is in the rural area.

[0025] In a preferred example, the present application may be further configured as follows: based on the real-time signaling data and the preset type data, the step of extracting the key features of the target user includes:

[0026] Performing analysis and mining based on the real-time signaling data and the preset type data, and statistically analyzing the user characteristics of tourist user samples and non-tourist user samples;

[0027] The key features of the target user are extracted from the user features of the tourist user sample, wherein the key features include having visited other scenic spots in the province, moving within a target distance range, the time spent at the scenic spot reaching a preset time threshold, the Internet traffic being lower than a preset traffic threshold, the number of calls on the day being lower than a preset number threshold, and the mobile phone package charges exceeding a preset charge threshold.

[0028] In a preferred example, the present application can be further configured as follows: it also includes the following steps:

[0029] Count the number of words in tourism marketing text messages;

[0030] When the number of words in the tourism marketing SMS exceeds a preset word count threshold, the tourism marketing SMS is cut according to the word count threshold to obtain a plurality of sub-short messages in sequence;

[0031] Send several sub-short messages to the tourist user terminal in batches according to the order of cutting.

[0032] In a second aspect, the present application provides a device for sending tourism marketing SMS messages.

[0033] This application is achieved through the following technical solutions:

[0034] A tourism marketing SMS sending device, comprising:

[0035] A data collection module, used to collect real-time signaling data of target users in a target scenic spot and other preset types of data from operators;

[0036] A feature extraction module, used to extract key features of a target user based on the real-time signaling data and the preset type data;

[0037] A tourist user identification module is used to input the key features into a preset LightGBM model for identification and output identification results of tourist users;

[0038] The SMS configuration module is used to match the sending number, SMS content, marketing area base station and sending time corresponding to the tourist user according to the identification result of the tourist user, and complete the pre-configuration of the tourist marketing SMS sending task;

[0039] The SMS sending module is used to trigger the tourism marketing SMS sending task and send tourism marketing SMS when the tourist user enters the target marketing area.

[0040] In a third aspect, the present application provides a computer device.

[0041] This application is achieved through the following technical solutions:

[0042] A computer device comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of any one of the above-mentioned tourism marketing SMS sending methods when executing the computer program.

[0043] In a fourth aspect, the present application provides a computer-readable storage medium.

[0044] This application is achieved through the following technical solutions:

[0045] A computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of any one of the above-mentioned tourism marketing SMS sending methods are implemented.

[0046] In a fifth aspect, the present application provides a computer program product.

[0047] This application is achieved through the following technical solutions:

[0048] A computer program product includes a computer program, which implements the steps of any one of the above-mentioned tourism marketing SMS sending methods when executed by a processor.

[0049] In summary, compared with the prior art, the technical solution provided by this application has at least the following beneficial effects:

[0050] Collect real-time signaling data of target users in target scenic spots and other preset type data from operators to more accurately reflect the behavior of target users; extract key features of target users based on real-time signaling data and preset type data as input data for model recognition; input key features into the preset LightGBM model for recognition, and output recognition results of tourist users, so as to achieve the purpose of building user portraits through AI algorithms and more accurately identifying tourist users, overcoming the problem of inaccuracy when identifying target users using resident users and manual rules; according to the recognition results of tourist users, match the sending number, SMS content, marketing area base station and sending time corresponding to the tourist users, complete the pre-configuration of the tourism marketing SMS sending task, and when the tourist user enters the target marketing area, trigger the tourism marketing SMS sending task and send the tourism marketing SMS, which improves the push accuracy of tourism marketing SMS and improves the SMS harassment problem for local residents, while greatly improving the efficiency of SMS sending and reducing the waste of tourism marketing SMS charges. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 A schematic diagram of the main process of a method for sending tourism marketing SMS is provided as an exemplary embodiment of the present application.

[0052] Figure 2 A schematic diagram of the configuration of a text message sending task of a tourism marketing text message sending method provided as yet another exemplary embodiment of the present application.

[0053] Figure 3 A schematic diagram of a method for sending tourism marketing text messages provided as another exemplary embodiment of the present application for mining and analyzing the characteristics of whether other scenic spots in the province have been visited.

[0054] Figure 4 A schematic diagram of mining and analyzing movement characteristics involving longer distances in a method for sending tourism marketing SMS provided as an exemplary embodiment of the present application.

[0055] Figure 5 A schematic diagram of a method for sending tourism marketing SMS provided as an exemplary embodiment of the present application for mining and analyzing whether there is a long-term stay feature at a scenic spot.

[0056] Figure 6 A schematic diagram of mining and analyzing whether a tourism marketing SMS sending method has high Internet traffic characteristics provided as an exemplary embodiment of the present application.

[0057] Figure 7 A schematic diagram of mining and analyzing calling number characteristics in a method for sending tourism marketing SMS provided as an exemplary embodiment of the present application.

[0058] Figure 8 A schematic diagram of mining and analyzing package tariff characteristics in a method for sending tourism marketing SMS provided as an exemplary embodiment of the present application.

[0059] Fig. 9 A schematic diagram of SMS content configuration for a method for sending tourism marketing SMS messages provided as an exemplary embodiment of the present application. DETAILED DESCRIPTION

[0060] This specific embodiment is merely an explanation of the present application and is not a limitation of the present application. After reading this specification, those skilled in the art may make modifications to the present embodiment without any creative contribution as needed, but such modifications are protected by the patent law as long as they are within the scope of the claims of the present application.

[0061] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0062] In addition, the term "and / or" in this article is only a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in this article, unless otherwise specified, generally means that the associated objects before and after are in an "or" relationship.

[0063] The embodiments of the present application are further described in detail below in conjunction with the drawings in the specification.

[0064] Reference Figure 1 , an embodiment of the present application provides a method for sending tourism marketing SMS, and the main steps of the method are described as follows.

[0065] S1: Collecting real-time signaling data of target users in the target scenic area and other preset type data from the operator;

[0066] S2: extracting key features of the target user based on the real-time signaling data and the preset type data;

[0067] S3: Input the key features into the preset LightGBM model for identification, and output the identification results of the tourist users;

[0068] S4: According to the identification result of the tourist user, the sending number, SMS content, marketing area base station and sending time corresponding to the tourist user are matched to complete the pre-configuration of the tourism marketing SMS sending task;

[0069] S5: When the tourist user enters the target marketing area, the tourist marketing SMS sending task is triggered and the tourist marketing SMS is sent.

[0070] Specifically, the latitude and longitude information of a specific scenic spot and the base station information are issued, and the base station information of the specific scenic spot is configured to monitor the target users roaming into the target scenic spot in real time.

[0071] Real-time signaling data of target users in the target scenic spot and other preset types of data from the operator are collected to construct the behavioral characteristics of tourist users, and then tourist users are identified by constructing a tourist user portrait model. This solves the problem that the existing technology often starts from the perspective of resident users and adopts artificial rules to identify target users. The target users are obtained by eliminating the resident users after identification, which has identification limitations and cannot accurately identify resident users, making it difficult to accurately screen out tourist users.

[0072] The real-time signaling data may be voice call signaling, location update signaling, attach signaling generated by Internet access, detach signaling generated by Internet access, and the like.

[0073] By parsing real-time signaling data, matching all base stations within the longitude and latitude of a specific scenic spot, and monitoring mobile users under the corresponding base stations, operators can collect other types of data from telecom operators in addition to real-time signaling data from target users in the target scenic spot for various key feature extraction of tourist users.

[0074] By using the extracted key features of the target users and building user portraits through AI machine learning models, the LightGBM model is obtained to identify tourist users. In this way, user portraits are built with the help of cutting-edge AI algorithms to overcome the limitations of existing technologies in identifying tourist users.

[0075] The LightGBM model used in this embodiment is a gradient boosting algorithm based on a decision tree. It has the advantages of supporting parallel processing, fast training speed, high accuracy, and small memory usage. It is an optimized version based on the gradient boosting tree algorithm.

[0076] Among them, Gradient Boosting is a boosting method in ensemble learning. It constructs a strong classifier by combining several weak classifiers. The core idea of ​​gradient boosting is to gradually optimize the model by approximating the residual by negative gradient. The negative gradient value of the model is used to approximate the residual value T of the key parameter of model optimization, so that each subsequent iteration of the model reduces the overall loss function value. Gradient boosting uses decision trees as the base classifier.

[0077] As a kind of gradient boosting algorithm, the most important improvement of the LightGBM model is the use of the histogram algorithm and the leaf-wise algorithm with depth limitation.

[0078] The histogram algorithm discretizes continuous floating-point features into K integers to construct a histogram with a width of K. When traversing the data, the statistics are accumulated in the histogram based on the discretized values ​​as indexes. After traversing the data once, the histogram accumulates the required statistics, and then the optimal segmentation point is traversed from the discrete values ​​of the histogram. Compared with precise segmentation, although there is a certain decrease in accuracy, since the weak classifier does not require high accuracy, this method has little effect in the gradient boosting framework and can effectively prevent model overfitting.

[0079] The traditional tree growth strategy uses a level-wise strategy, but many leaves have low split gains and are not necessary to grow. Therefore, LightGBM uses a leaf-wise strategy: each time it grows from the node with the largest delta loss among all the current leaves. This growth method needs to be used with the parameters of the maximum depth limit (max_depth) and the number of leaves (number_leaves) to ensure high efficiency while limiting the tree from being too deep or too many leaves to prevent the model from overfitting.

[0080] Therefore, LightGBM supports parallel processing, fast training speed, high accuracy, and small memory usage, making it very suitable for predicting tourist users in the scenario of sending tourism marketing SMS.

[0081] Then the key features are input into the above-mentioned LightGBM model for identification, and the identification results of tourist users are output.

[0082] According to the identification results of tourist users, the SMS center function is called to match the sending number, SMS content, marketing area base station and sending time corresponding to the tourist user, and complete the pre-configuration of the tourism marketing SMS sending task, such as Figure 2 shown.

[0083] The system consumes the registration signaling data in the open source stream processing platform Kafka in real time to determine whether the tourist user has entered the marketing area. The open source stream processing platform Kafka is a high-throughput distributed publish-subscribe messaging system that can quickly process stream data and improve data processing efficiency.

[0084] When a tourist user enters the target marketing area, the SMS center is connected to trigger the tourist marketing SMS sending task, and the tourist marketing SMS is sent to achieve accurate sending of the tourist marketing SMS, thereby avoiding the problem of inaccurate identification of target users using resident users and manual rules, improving the SMS harassment problem for local resident people, and greatly improving the efficiency of SMS sending, avoiding the waste of tourist marketing SMS charges.

[0085] In one embodiment, the real-time signaling data includes VOLTE registration signaling and call signaling, the preset type data includes Internet bill data, user package value data, whether the package is integrated, whether the ID card data is from another place, and whether the registered residence is in a rural area. When collecting the real-time signaling data of the target user located in the target scenic spot and the rest of the preset type data from the operator, it includes:

[0086] Collect VOLTE registration signaling and call signaling of target users in the target scenic area; and

[0087] Collect Internet call data from operators, user package value data, whether the package is integrated, whether the ID card is from another place, and whether the registered residence is in a rural area. The user package value, whether the package is integrated, whether the ID card is from another place, and whether the registered residence is in a rural area constitute the user's basic data.

[0088] Among them, the key fields of VOLTE registration signaling are shown in Table 1 below.

[0089] Table 1

[0090]

[0091] The key fields of call signaling are shown in Table 2 below.

[0092] Table 2

[0093]

[0094] The key fields of the Internet call bill data are shown in Table 3 below.

[0095] Table 3

[0096]

[0097]

[0098] By collecting VOLTE registration signaling, call signaling, Internet bill data, user package value data, whether the package is integrated, whether the ID card data is from another place, and whether the household registration place is in the rural area of ​​the target users located in the target scenic spot, we can better portray the user portrait of the target users, which is conducive to more accurate identification of tourist users in the future.

[0099] In one embodiment, the step of extracting key features of the target user based on the real-time signaling data and the preset type data includes:

[0100] Performing analysis and mining based on the real-time signaling data and the preset type data, and statistically analyzing the user characteristics of tourist user samples and non-tourist user samples;

[0101] The key features of the target user are extracted from the user features of the tourist user sample, wherein the key features include having visited other scenic spots in the province, moving within a target distance range, the time spent at the scenic spot reaching a preset time threshold, the Internet traffic being lower than a preset traffic threshold, the number of calls on the day being lower than a preset number threshold, and the mobile phone package charges exceeding a preset charge threshold.

[0102] In this embodiment, the tourist user list compiled by a provincial cultural and tourism bureau is used as the analysis sample source of tourist users, and the local users compiled by communities and other departments are used as the analysis sample source of non-tourist users, including 64,500 non-tourist users and 13,120 tourist users, totaling 77,620 numbers.

[0103] A comparative analysis is conducted on the characteristics of the two types of users to uncover the typical characteristics of tourist users, that is, the characteristics that tourist users have but non-tourist users do not have.

[0104] Through the behavioral analysis reflected by the VOLTE registration signaling, call signaling, Internet bill data, user package value data, whether the package is integrated data, whether the ID card data is from another place and whether the registered residence is in a rural area of ​​the tourist user number, a focused analysis is conducted based on the preset key dimension features in order to discover the characteristics that can distinguish tourist users from non-tourist users.

[0105] Key dimension features include whether the person has visited other scenic spots in the province, such as Figure 3 shown.

[0106] from Figure 3 It can be seen that most of the non-tourism users have not visited other scenic spots in the province, while most of the tourism users have visited other scenic spots in the province. This is also in line with reality, because there are many scenic spots in a province, and currently tourism users tend to visit multiple scenic spots instead of just one of them.

[0107] Key dimension features include whether it involves long-distance movement features, such as Figure 4 shown.

[0108] from Figure 4 It can be seen that most of the non-tourism users do not move long distances, while most of the tourism users do. This is also in line with reality, because tourism users often take high-speed trains, airplanes, cars and other means of transportation to move long distances, while non-tourism users often do not move long distances unless they happen to be on a business trip or other special scenarios. Among them, the long distance can be more than 30 kilometers.

[0109] Key dimension features include whether there is a long stay at the scenic spot, such as Figure 5 shown.

[0110] from Figure 5 It can be seen that most of the non-tourism users do not stay in scenic spots for a long time, while most of the tourism users stay in scenic spots for a long time. This is also in line with reality, because tourism users need to visit scenic spots and stay for a long time, while non-tourism users are not in the scenic area or passing through the scenic area, and basically have no stay time. Among them, a long time can be more than 1 hour.

[0111] Key dimension features include whether there is high Internet traffic characteristics, such as Figure 6 shown.

[0112] from Figure 6 It can be seen that most non-tourism users have high Internet traffic, while most tourism users do not have high Internet traffic. This is also in line with reality, because non-tourism users often watch short videos in their daily lives, which normally generates a certain amount of Internet traffic every day. Tourism users are often busy traveling and do not generate high Internet traffic. Therefore, such features can be extracted based on the operator's Internet call bill for distinction. If the user's Internet traffic on that day exceeds 2G, it is determined that the user has high Internet traffic; if the user's Internet traffic on that day is less than 1G, it is determined that the user does not have high Internet traffic.

[0113] Key dimension features include calling times features, such as Figure 7 shown.

[0114] from Figure 7 It can be seen that most of the non-tourist users have high outbound calls on the same day, while most of the tourist users have low outbound calls on the same day. This is also in line with reality, because non-tourist users often need to make outbound calls due to work reasons, and tourist users are mainly for travel and do not need too many outbound calls. Among them, you can set whether the daily call volume is greater than 10 times to measure the call frequency characteristics. If the daily call volume is greater than 10 times, it is determined that the user has high outbound calls on the same day, otherwise it is determined that the user has low outbound calls on the same day.

[0115] Key dimension features include package fee features, such as Figure 8 shown.

[0116] from Figure 8 It can be seen that most non-travel users have lower package charges, while most travel users have higher package charges. This is also in line with reality, because most travel users are more affluent than non-travel users, and their package charges are correspondingly higher. Among them, if the user's package charge exceeds 139 yuan, it is determined that the user's package charge is higher.

[0117] Through the above analysis, based on real-time signaling data and preset type data, the key features of the target users are extracted as shown in Table 4 below.

[0118] Table 4

[0119]

[0120]

[0121] In one embodiment, the steps of constructing the LightGBM model include:

[0122] Initialize the training parameters of the LightGBM model, including the number of trees n_estimators, the maximum depth of the tree max_depth, the learning rate learning_rate, the number of leaf nodes num_leaves of the tree, and the class weight ratio class_weights;

[0123] Input historical key features into the initialized LightGBM model for learning, evaluate the current AUC value and KS value of the model, and optimize the training parameters of the current LightGBM model based on the current AUC value and KS value;

[0124] When the current AUC value and KS value of the model meet the preset conditions, the current LightGBM model is output as the preset LightGBM model.

[0125] In one embodiment, the preset training parameters of the LightGBM model meet the following conditions:

[0126] The number of trees n_estimators = 1000;

[0127] The maximum depth of the tree max_depth = 5;

[0128] Learning rate learning_rate = 0.27;

[0129] The number of leaf nodes in the tree num_leaves = 61;

[0130] Class weight ratio class_weights = [1,300].

[0131] Specifically, based on the LightGBM model, the parameters of the LightGBM model are adjusted and optimized. The specific steps are as follows:

[0132] ① Initial parameter setting;

[0133] First, a set of basic parameters are set for the initial training of the model, including the number of trees n_estimators = 1000, the maximum depth of the tree max_depth = 5, the learning rate learning_rate = 0.27, and the number of leaf nodes num_leaves = 61. The initial selection of these basic parameters is based on previous experience and the results of preliminary experiments.

[0134] ②Parameter optimization;

[0135] Adjust the number of leaf nodes num_leaves and class weight ratio class_weights: Experiment with the number of num_leaves and different class_weights to explore the sensitivity of the model to imbalanced classes. For example, try setting num_leaves to 128 and different class weight ratios, such as [1,150] to [1,300], to observe the changes in AUC and KS values.

[0136] AUC (Area Under the Curve) and KS value (Kolmogorov-Smirnov) are two important indicators for evaluating the performance of binary classification models. They measure the pros and cons of the model from different perspectives.

[0137] AUC is the area under the ROC curve (Receiver Operating Characteristic Curve), which is used to measure the model's ability to distinguish between positive and negative samples. The ROC curve uses the false positive rate (FPR) as the horizontal axis and the true positive rate (TPR) as the vertical axis. The larger the AUC value, the stronger the model's ability to distinguish between positive and negative samples. Usually, the AUC value is between 0.5 and 1. The closer the AUC value is to 1, the better the model performance. The higher the AUC value, the stronger the generalization ability of the model.

[0138] The KS value is a measure of the model's ability to distinguish between good and bad samples and is usually used in credit scoring models. The KS value is defined as the difference between the maximized TPR and FPR, that is, the maximum value of the difference between the cumulative positive sample rate (TPR) and the cumulative negative sample rate (FPR). The larger the KS value, the stronger the model's ability to distinguish between good and bad samples. Usually, a KS value greater than 0.35 is considered to be a good model performance. The higher the KS value, the stronger the model's ability to distinguish.

[0139] AUC can help us understand the overall performance of the model, and the KS value can quantify the model's ability to distinguish customers with different credit ratings. Models with high KS values ​​can more accurately identify high-risk and low-risk customers and improve recognition accuracy.

[0140] Also, adjust learning_rate and max_depth: By fine-tuning these key parameters, try to find the optimal model complexity and learning speed. For example, adjust learning_rate from 0.3 to 0.4 or 0.2, and test the impact on model performance.

[0141] Model evaluation: After each parameter adjustment, the same training set and validation set are used to evaluate the AUC and KS values ​​of the model to help understand the model performance effects under different parameter configurations.

[0142] ③ Final parameters are determined;

[0143] After multiple rounds of parameter adjustment and model evaluation, it was found that when n_estimators = 1000, max_depth = 5, learning_rate = 0.27, num_leaves = 61, and class_weights = [1,300] were set, the LightGBM model achieved a relatively balanced optimal result in terms of accuracy and recall. At this time, the accuracy of the LightGBM model reached 96.18% and the recall rate reached 95.53%, as shown in Table 5 below:

[0144] Table 5

[0145]

[0146] The LightGBM model is optimized through the above process, and then the key features are input into the optimized LightGBM model for identification, which can accurately identify tourist users, achieve high accuracy and recall rates, and ensure the robustness and efficiency of the model when processing unbalanced data.

[0147] In one embodiment, before learning historical key features, positive and negative samples are labeled.

[0148] Specifically, by selecting analysis samples, such as using the list of tourist users counted by a provincial cultural and tourism bureau as the source of analysis samples for tourist users, and using local users counted by communities and other departments as the source of analysis samples for non-tourism users, positive and negative sample labeling is performed for the characteristics of the two types of users. Among them, there are 64,500 non-tourism users and 13,120 tourist users, totaling 77,620 numbers, and the data set is divided into two parts: a training set and a test set. The training set is used for model training and parameter adjustment, and the test set is used for the final model evaluation. In this embodiment, the training set accounts for 80% of the data set, and the test set accounts for 20% of the data set. At the same time, some features are binarized, with 0 representing no and 1 representing yes. The positive and negative sample labeling of the data set is shown in Table 6 below.

[0149] Table 6

[0150]

[0151] Finally, the historical key features after positive and negative samples are labeled are input into the LightGBM model for learning. By preprocessing the input data, the model can complete the recognition work faster and with higher recognition efficiency.

[0152] Historical key features may be historical key features such as target users who have visited other scenic spots in the province, moved within the target distance range, stayed at the scenic spot for a preset time threshold, had Internet traffic lower than a preset traffic threshold, had the number of calls on the day lower than a preset number threshold, and had their mobile phone package charges exceed a preset charge threshold.

[0153] In one embodiment, a method for sending tourism marketing SMS also includes the following steps:

[0154] Count the number of words in tourism marketing text messages;

[0155] When the number of words in the tourism marketing SMS exceeds a preset word count threshold, the tourism marketing SMS is cut according to the word count threshold to obtain a plurality of sub-short messages in sequence;

[0156] Send several sub-short messages to the tourist user terminal in batches according to the order of cutting.

[0157] Reference Fig. 9 , by flexibly configuring the SMS content, long SMS can be automatically split into short SMS for easy billing. Among them, long SMS refers to tourism marketing SMS with a word count exceeding the word count threshold. Short SMS refers to tourism marketing SMS with a word count less than or equal to the word count threshold. The tourism user terminal can support the automatic merging of several segmented sub-SMS into a long SMS.

[0158] Because the cultural and tourism departments of all provinces have proposed the demand for sending precise tourism marketing text messages, taking a certain city as an example, during holidays, welcome text messages are sent to tourists visiting from other places within the city. The contract amount is 100,000 yuan, and the settlement is 0.04 yuan per text message. Therefore, practice has proved that a method of sending tourism marketing text messages has high commercial value.

[0159] In summary, a method for sending tourism marketing SMS collects real-time signaling data of target users located in a target scenic spot and other preset type data from operators to more accurately reflect the behavior of target users; based on real-time signaling data and preset type data, the key features of the target users are extracted as input data for model recognition; the key features are input into the preset LightGBM model for recognition, and the recognition results of the tourist users are output, so as to achieve the purpose of building user portraits through AI algorithms and more accurately identifying tourist users, and overcome the problem of inaccuracy when identifying target users using resident users and manual rules; according to the recognition results of tourist users, the sending number, SMS content, marketing area base station and sending time corresponding to the tourist users are matched to complete the pre-configuration of the tourism marketing SMS sending task, and when the tourist user enters the target marketing area, the tourism marketing SMS sending task is triggered, and the tourism marketing SMS is sent, which can improve the evaluation accuracy of tourist users, improve the push accuracy of tourism marketing SMS, and improve the SMS harassment problem of local resident people, while greatly improving the efficiency of SMS sending and reducing the waste of tourism marketing SMS charges.

[0160] A method for sending tourism marketing text messages uses a machine learning algorithm to learn the key features, thresholds, weights, etc. used to calculate the user portrait of the constructed target user, which is objective and avoids the limitations brought by the resident user filtering method and the artificial rules constructed by human subjective cognition. It can more accurately identify the tourist population and greatly improve the accuracy of tourist user identification.

[0161] At the same time, the LightGBM model supports parallel processing, fast training speed, high accuracy, and small memory usage, making it very suitable for predicting tourist users in the scenario of sending tourism marketing SMS. The optimized LightGBM model can accurately identify tourist users, achieve high accuracy and recall rates, and ensure the robustness and efficiency of the model when processing unbalanced data.

[0162] It should be understood that the size of the serial numbers of the steps in the above embodiments does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0163] The present application also provides a tourism marketing SMS sending device, which corresponds to a tourism marketing SMS sending method in the above embodiment. The tourism marketing SMS sending device includes:

[0164] A data collection module, used to collect real-time signaling data of target users in a target scenic spot and other preset types of data from operators;

[0165] A feature extraction module, used to extract key features of a target user based on the real-time signaling data and the preset type data;

[0166] A tourist user identification module is used to input the key features into a preset LightGBM model for identification and output identification results of tourist users;

[0167] The SMS configuration module is used to match the sending number, SMS content, marketing area base station and sending time corresponding to the tourist user according to the identification result of the tourist user, and complete the pre-configuration of the tourist marketing SMS sending task;

[0168] The SMS sending module is used to trigger the tourism marketing SMS sending task and send tourism marketing SMS when the tourist user enters the target marketing area.

[0169] For the specific limitations of a device for sending tourism marketing SMS, please refer to the limitations of a method for sending tourism marketing SMS in the above text, which will not be repeated here.

[0170] Each module in the above-mentioned tourism marketing SMS sending device can be implemented in whole or in part by software, hardware or a combination thereof. Each module can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute the operations corresponding to each module.

[0171] In one embodiment, a computer device is provided, which may be a server. The computer device includes a processor, a memory, a network interface, and a database connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, any one of the above-mentioned tourism marketing SMS sending methods is implemented.

[0172] In one embodiment, a computer-readable storage medium is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, and when the processor executes the computer program, any one of the above-mentioned tourism marketing SMS sending methods is implemented.

[0173] In one embodiment, a computer program product is provided, which includes a computer program, and when the computer program is executed by a processor, any one of the above-mentioned tourism marketing SMS sending methods is implemented.

[0174] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiments can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium, including several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present application. When the computer program is executed, it may include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in the present application may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0175] Those skilled in the art will clearly understand that for the sake of convenience and brevity of description, only the division of the above-mentioned functional units and modules is used as an example. In actual applications, the above-mentioned functions can be distributed and completed by different functional units and modules as needed, that is, the internal structure of the system can be divided into different functional units or modules to complete all or part of the functions described above.

Claims

1. A method for sending tourism marketing SMS, characterized in that: The following steps are included: Collect real-time signaling data of target users in target scenic spots and other preset data from operators; Extracting key features of the target user based on the real-time signaling data and the preset type data; Input the key features into the preset LightGBM model for identification, and output the identification results of the tourist users; According to the identification result of the tourist user, the sending number, SMS content, marketing area base station and sending time corresponding to the tourist user are matched to complete the pre-configuration of the tourist marketing SMS sending task; When a tourist user enters a target marketing area, the tourist marketing SMS sending task is triggered, and a tourist marketing SMS is sent.

2. The method for sending tourism marketing SMS according to claim 1, characterized in that: The construction steps of the LightGBM model include: Initialize the training parameters of the LightGBM model, including the number of trees n_estimators, the maximum depth of the tree max_depth, the learning rate learning_rate, the number of leaf nodes num_leaves of the tree, and the class weight ratio class_weights; Input historical key features into the initialized LightGBM model for learning, evaluate the current AUC value and KS value of the model, and optimize the training parameters of the current LightGBM model based on the current AUC value and KS value; When the current AUC value and KS value of the model meet the preset conditions, the current LightGBM model is output as the preset LightGBM model.

3. The method for sending tourism marketing SMS according to claim 2, characterized in that: The preset training parameters of the LightGBM model meet the following conditions: The number of trees n_estimators = 1000; The maximum depth of the tree max_depth = 5; Learning rate learning_rate = 0.27; The number of leaf nodes in the tree num_leaves = 61; Class weight ratio class_weights = [1,300].

4. The method for sending tourism marketing SMS according to claim 1, characterized in that: The real-time signaling data includes VOLTE registration signaling and call signaling, the preset type data includes Internet bill data, user package value data, whether it is integrated package data, whether it is a foreign ID card data and whether the registered residence is rural data, and the real-time signaling data of the target user located in the target scenic spot and the rest of the preset type data from the operator are collected, including: Collect VOLTE registration signaling and call signaling of target users in the target scenic area; as well as Collect Internet call bill data from operators, user package value data, whether the package is integrated, whether the ID card data is from another place, and whether the household registration place is in the rural area.

5. The method for sending tourism marketing SMS according to claim 1, characterized in that: Based on the real-time signaling data and the preset type data, the step of extracting key features of the target user includes: Performing analysis and mining based on the real-time signaling data and the preset type data, and statistically analyzing the user characteristics of tourist user samples and non-tourist user samples; The key features of the target user are extracted from the user features of the tourist user sample, wherein the key features include having visited other scenic spots in the province, moving within a target distance range, the time spent at the scenic spot reaching a preset time threshold, the Internet traffic being lower than a preset traffic threshold, the number of calls on the day being lower than a preset number threshold, and the mobile phone package charges exceeding a preset charge threshold.

6. The method for sending tourism marketing SMS according to any one of claims 1 to 5, characterized in that: The following steps are also included: Count the number of words in tourism marketing text messages; When the number of words in the tourism marketing SMS exceeds a preset word count threshold, the tourism marketing SMS is cut according to the word count threshold to obtain a plurality of sub-short messages in sequence; Send several sub-short messages to the tourist user terminal in batches according to the order of cutting.

7. A tourism marketing SMS sending device, characterized in that: include, A data collection module, used to collect real-time signaling data of target users in a target scenic spot and other preset types of data from operators; A feature extraction module, used to extract key features of a target user based on the real-time signaling data and the preset type data; A tourist user identification module is used to input the key features into a preset LightGBM model for identification and output identification results of tourist users; The SMS configuration module is used to match the sending number, SMS content, marketing area base station and sending time corresponding to the tourist user according to the identification result of the tourist user, and complete the pre-configuration of the tourist marketing SMS sending task; The SMS sending module is used to trigger the tourism marketing SMS sending task and send tourism marketing SMS when the tourist user enters the target marketing area.

8. A computer device, characterized in that: The method comprises a memory, a processor and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the method according to any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

10. A computer program product, characterized in that The invention comprises a computer program, which implements the steps of the method according to any one of claims 1 to 6 when being executed by a processor.

Images