Method and device for predicting communication service cost and electronic equipment

Abstract

The invention provides a communication service cost prediction method and device and electronic equipment. The method comprises the following steps: acquiring opened service data and user information of a target building in a historical time period, and generating a user portrait according to the opened service data and the user information; obtaining non-opened service data of the target building, generating a decision tree model corresponding to the non-opened service data through the user portrait, and predicting first income data corresponding to the non-opened service data in a future time period through the decision tree model; and analyzing the opened service data through a triple exponential smoothing model, predicting to obtain second income data corresponding to the opened service data in a future time period, calculating the sum of the first income data and the second income data to obtain total income data in the future time period, and determining a communication operation plan according to the total income data and performing operation. The problem that the corresponding investment and operation plan is not accurate due to the fact that the error of communication income and cost prediction of the commercial building is large is solved.

Description

Technical Field

[0001] The present application relates to the technical field of communication data prediction, and in particular to a method, device, computer-readable storage medium and electronic device for predicting communication service fees. Background Art

[0002] The resource activation, construction planning, investment and operation of commercial buildings require analysis and decision-making based on past and future commercial building income data. With the rapid development of data analysis and deduction technology, it is necessary not only to view and display the current status of various types of data of commercial buildings, but also to conduct data analysis through effective future deduction to predict the future development trend of commercial buildings, so as to guide precise investment in planning, construction, customer operation, etc. However, the existing deduction methods have the following deficiencies when processing data with complex characteristics:

[0003] Limitations of traditional time series analysis methods: Existing time series models (such as ARIMA) can usually only process single-dimensional time series data, and are not accurate enough when dealing with buildings with multi-dimensional characteristics and no historical revenue data. For example, for commercial buildings that have not opened business types, it is impossible to deduce through historical revenue data.

[0004] Limitations of a single machine learning model: Although machine learning algorithms (such as a single decision tree, linear regression, etc.) have been used in income deduction, these models are often unable to handle complex feature interactions and are prone to overfitting or underfitting problems, resulting in less than ideal deduction results.

[0005] Imperfect feature processing methods: Traditional methods often ignore the interactions between features when processing multi-dimensional business situation features, resulting in the model being unable to fully capture the potential drivers of revenue changes.

[0006] Therefore, when faced with complex multi-dimensional business situation characteristics and historical revenue data, existing technologies are unable to provide sufficient deduction accuracy and robustness, resulting in large errors in commercial building revenue deduction, affecting the accuracy of the overall decision. Summary of the invention

[0007] The main purpose of the present application is to provide a method, device, computer-readable storage medium and electronic device for predicting communication service costs, so as to at least solve the problem in the prior art that the communication income and cost prediction of commercial buildings has large errors, resulting in inaccurate corresponding investment and operation plans.

[0008] In order to achieve the above-mentioned purpose, according to one aspect of the present application, a method for predicting communication service fees is provided, comprising: obtaining activated service data and user information of a target building in a historical time period, and generating a user portrait based on the activated service data and the user information, wherein the activated service data represents the communication services that the user has activated, the user information represents the identity and characteristics of the user, and the user portrait is a virtual image that at least characterizes the characteristics of the user; obtaining unactivated service data of the target building, generating a decision tree model corresponding to the unactivated service data through the user portrait, and predicting future The first revenue data corresponding to the unactivated service data of a time period, wherein the decision tree model includes multiple sub-nodes, and the unactivated service data represents the communication services that the user has not activated; the activated service data is analyzed by a triple exponential smoothing model to predict the second revenue data corresponding to the activated service data of the future time period, and the sum of the first revenue data and the second revenue data is calculated to obtain the total revenue data of the future time period; the communication operation plan of the target building is determined according to the total revenue data and is operated, wherein the communication operation plan represents the plan for the communication services of the future time period.

[0009] Optionally, generating a decision tree model corresponding to the unactivated service data through the user portrait includes: obtaining user features in the user portrait, and taking each data in the unactivated service data as a sub-data set, and correcting the sub-data set according to the user features to obtain the corrected sub-data set; for each of the sub-data sets, constructing the decision tree model using an unpruned decision tree algorithm, wherein the decision tree model is not pruned during the construction process, and only one feature is selected for judgment when each node is split.

[0010] Optionally, the activated service data is analyzed through a triple exponential smoothing model to predict second revenue data corresponding to the activated service data in the future time period, including: establishing a time series index for the activated service data; analyzing and adjusting the activated service data and the corresponding time series index through a triple exponential smoothing model to obtain smoothing parameters of the triple exponential smoothing model, wherein the smoothing parameters are parameters used to characterize the degree of change of the activated service data over time; and predicting the revenue of the activated service data in the future time period through a triple exponential smoothing model configured with the smoothing parameters to obtain the second revenue data.

[0011] Optionally, after obtaining the activated service data of the target building in the historical time period and obtaining the unactivated service data of the target building, the method further includes: standardizing the activated service data and the unactivated service data, respectively; encoding the activated service data and the unactivated service data after standardization to obtain first coded data corresponding to the activated service data and second coded data corresponding to the unactivated service data.

[0012] Optionally, a user portrait is generated based on the activated service data and the user information, including: extracting key features that have a significant impact on user behavior in the activated service data; dividing the users into different groups by statistical methods, and analyzing the behavior patterns of each of the groups; and constructing the user portrait by combining the key features and the behavior patterns.

[0013] Optionally, after obtaining the total revenue data for the future time period, the method further includes: obtaining actual revenue data; calculating error parameters between the actual revenue data and the total revenue data to evaluate the prediction accuracy, wherein the error parameter is a parameter that characterizes the error size.

[0014] Optionally, obtaining the activated service data and user information of the target building in a historical time period includes: obtaining the user information and the activated service data of the user by matching the service number, wherein the service number is a unique number generated by the user in the process of activating the service.

[0015] According to another aspect of the present application, a communication service fee prediction device is provided, comprising: a generation unit, used to obtain activated service data and user information of a target building in a historical time period, and generate a user portrait based on the activated service data and the user information, wherein the activated service data represents the communication service that the user has activated, the user information represents the identity and characteristics of the user, and the user portrait is a virtual image that characterizes the characteristics of the user; a prediction unit, used to obtain unactivated service data of the target building, generate a decision tree model corresponding to the unactivated service data through the user portrait, and predict the future time through the decision tree model The decision tree model includes a plurality of sub-nodes, and the unactivated service data represents the communication services that the user has not activated; a calculation unit is used to analyze the activated service data through a triple exponential smoothing model, predict and obtain the second income data corresponding to the activated service data in the future time period, calculate the sum of the first income data and the second income data, and obtain the total income data of the future time period; determine the communication operation plan of the target building according to the total income data and operate it, wherein the communication operation plan represents the plan for the communication service in the future time period.

[0016] According to another aspect of the present application, a computer-readable storage medium is provided, wherein the computer-readable storage medium includes a stored program, wherein when the program is executed, the device where the computer-readable storage medium is located is controlled to execute any one of the communication service cost prediction methods.

[0017] According to another aspect of the present application, an electronic device is provided, comprising: one or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and are configured to be executed by the one or more processors, and the one or more programs include a method for predicting the communication service cost for execution of any one of the described methods.

[0018] By applying the technical solution of the present application, the activated service data and user information of the target building in the historical time period are obtained, and a user portrait is generated based on the activated service data and user information; the unactivated service data of the target building is obtained, and a decision tree model corresponding to the unactivated service data is generated through the user portrait, and the first income data corresponding to the unactivated service data in the future time period is predicted through the decision tree model; the activated service data is analyzed through the triple exponential smoothing model, and the second income data corresponding to the activated service data in the future time period is predicted, and the sum of the first income data and the second income data is calculated to obtain the total income data for the future time period. Compared with the prior art, when faced with complex multi-dimensional business situation characteristics and historical income data, the error of commercial building income prediction is large, resulting in inaccurate corresponding investment and operation plans. In this application, by using different models to predict the user's activated service data and unactivated service data, the corresponding income can be accurately predicted according to the characteristics of the above data, so as to accurately determine and adjust the subsequent communication operation plan. Therefore, it is possible to solve the problems of large errors in predicting communication service costs and inaccurate communication operation planning in the prior art, and achieve the effect of accurately predicting communication service costs and accurately planning operation plans. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The drawings constituting part of the present application are used to provide a further understanding of the present application. The exemplary embodiments and descriptions of the present application are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0020] Figure 1 A hardware structure block diagram of a mobile terminal for executing a method for predicting communication service fees provided in an embodiment of the present application is shown;

[0021] Figure 2 A flow chart showing a method for predicting communication service fees provided in an embodiment of the present application is shown;

[0022] Figure 3 A specific communication service fee prediction method provided by an embodiment of the present application is shown;

[0023] Figure 4 A specific communication service fee prediction and deduction main line logic diagram provided by an embodiment of the present application is shown;

[0024] Figure 5 A schematic diagram of a decision tree provided in an embodiment of the present application is shown;

[0025] Figure 6 A flow chart of a deduction model provided by an embodiment of the present application is shown;

[0026] Figure 7A line chart of building income deduction provided by an embodiment of the present application is shown;

[0027] Figure 8 A flow chart showing a combination of a Holt-Winters triple exponential smoothing model and a random forest algorithm provided in an embodiment of the present application is shown;

[0028] Fig. 9 A structural block diagram of a communication service fee prediction device provided in an embodiment of the present application is shown.

[0029] The above drawings include the following reference numerals:

[0030] 102, processor; 104, memory; 106, transmission device; 108, input and output devices. DETAILED DESCRIPTION

[0031] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0032] In order to enable those skilled in the art to better understand the solution of the present application, 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 only 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 should fall within the scope of protection of the present application.

[0033] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequential order. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present application described here. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0034] For the convenience of description, some nouns or terms involved in the embodiments of the present application are explained below:

[0035] Commercial Building Revenue: The telecommunications operator calculates the total monthly communication service fees generated by all businesses of users of commercial buildings (such as personal fixed-line / broadband, government and enterprise dedicated lines, etc.).

[0036] Random Forest: A machine learning algorithm based on ensemble learning. By building multiple decision trees and integrating their results, random forests can improve the accuracy and robustness of predictions. Each decision tree is trained by randomly selecting features and no pruning is performed, allowing the model to handle complex feature interactions.

[0037] Decision Tree: A tree-structured model in which each node represents the splitting condition of a feature and the leaf node represents the predicted result. The decision tree is trained by recursively splitting the data set.

[0038] Random Feature Selection Mechanism: When constructing a random forest, in order to avoid overfitting, each node randomly selects some features instead of all features for judgment when splitting. This mechanism enables the model to have better generalization ability and avoids model bias caused by over-reliance on certain features.

[0039] Unpruned Decision Trees: Decision trees in random forests are usually not pruned, allowing the tree to grow to its maximum depth as much as possible. This helps capture complex patterns and feature interactions in the data, especially for high-dimensional data.

[0040] Triple Exponential Smoothing: The core algorithm in the Holt-Winters model, used to capture long-term trends, seasonal fluctuations, and current levels in time series data. This method is suitable for time series data with obvious trends and seasonal changes. Triple Exponential Smoothing predicts income trends and fluctuations by adjusting three smoothing parameters (α, β, γ).

[0041] Holt-Winters model: A statistical model for time series forecasting, suitable for revenue forecasting scenarios with trend and seasonal fluctuations. The model captures the law of revenue changes through three core parameters: level, trend, and seasonality, and is suitable for revenue forecasting of already leased orders.

[0042] Mean Squared Error (MSE): A model evaluation metric that measures the average sum of squared errors between predicted values ​​and actual values. The smaller the MSE, the more accurate the model prediction.

[0043] Root Mean Squared Error (RMSE): The square root of MSE, which is used to measure the standard deviation of the error between the predicted value and the actual value. RMSE is a variant of MSE and more intuitively reflects the magnitude of the error.

[0044] Mean Absolute Error (MAE): Another metric for evaluating the performance of a forecasting model that calculates the average absolute difference between the predicted values ​​and the actual values.

[0045] Coefficient of Determination: R 2 It indicates the model's ability to explain the data variance, ranging from 0 to 1, R 2 The closer it is to 1, the better the model fits the data.

[0046] One-Hot Encoding: A method of encoding categorical features into numerical form. Each category is represented by binary 0 and 1, which is very commonly used when processing categorical features.

[0047] As introduced in the background technology, in the prior art, when faced with complex multi-dimensional business situation characteristics and historical revenue data, the error in the communication revenue forecast of commercial buildings is large, resulting in inaccurate corresponding investment and operation plans. In order to solve the problem that the error in the communication revenue forecast of commercial buildings is large, resulting in inaccurate corresponding investment and operation plans, the embodiments of the present application provide a communication service fee prediction method, device, computer-readable storage medium and electronic device.

[0048] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0049] The method embodiments provided in the embodiments of the present application can be executed in a mobile terminal, a computer terminal or a similar computing device. Taking running on a mobile terminal as an example, Figure 1 FIG. 1 is a hardware structure block diagram of a mobile terminal of a method for predicting communication service fees according to an embodiment of the present invention. Figure 1 As shown, the mobile terminal may include one or more ( Figure 1 Only one is shown in the figure) a processor 102 (the processor 102 may include but is not limited to a processing device such as a microprocessor MCU or a programmable logic device FPGA) and a memory 104 for storing data, wherein the mobile terminal may also include a transmission device 106 and an input / output device 108 for communication functions. It can be understood by those skilled in the art that Figure 1The structure shown is only for illustration and does not limit the structure of the mobile terminal. Figure 1 More or fewer components as shown, or with Figure 1 Different configurations shown.

[0050] The memory 104 can be used to store computer programs, for example, software programs and modules of application software, such as the computer program corresponding to the method for predicting the communication service fee in the embodiment of the present invention. The processor 102 executes various functional applications and data processing by running the computer program stored in the memory 104, that is, the above method is implemented. The memory 104 may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some examples, the memory 104 may further include a memory remotely arranged relative to the processor 102, and these remote memories can be connected to the mobile terminal via a network. Examples of the above-mentioned network include but are not limited to the Internet, an intranet, a local area network, a mobile communication network, and a combination thereof. The transmission device 106 is used to receive or send data via a network. The specific example of the above-mentioned network may include a wireless network provided by a communication provider of the mobile terminal. In one example, the transmission device 106 includes a network adapter (Network Interface Controller, referred to as NIC), which can be connected to other network devices through a base station so as to communicate with the Internet. In one example, the transmission device 106 may be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0051] In this embodiment, a method for predicting communication service costs running on a mobile terminal, a computer terminal or a similar computing device is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0052] Figure 2 FIG. 1 is a flow chart of a method for predicting communication service fees according to an embodiment of the present application. Figure 2 As shown, the method comprises the following steps:

[0053] Step S201, obtaining the activated service data and user information of the target building in the historical time period, and generating a user portrait based on the activated service data and the user information, wherein the activated service data indicates the communication service that the user has activated, the user information indicates the identity and characteristics of the user, and the user portrait is a virtual image that at least represents the characteristics of the user;

[0054] Specifically, this real-time example uses buildings as units to count the communication costs of a building. The building being counted is the target building, and the activated service data and corresponding user information of users in the target building in the historical time period are extracted from the service order data. The activated service data includes but is not limited to: service type, package tariff, circuit level, rental duration, etc. At the same time, the collected user information, such as customer type, usage habits, preferences, etc., can be obtained from user registration information, market research or user feedback. Based on the above information, a user portrait of the user can be constructed. The user portrait includes a variety of user portrait features, including but not limited to user preferences, usage time period, customer type, and device type readiness information. The above user portrait features are obtained by in-depth analysis and summary of the above activated service data and user information, which not only characterizes the identity and characteristics of the user, but also reflects the user's actual demand and potential value for the building communication service. The generation and optimization of user portraits can not only help predict the communication service revenue of the building, but also guide the market strategy and operation decision of the building, such as optimizing services, adjusting packages, increasing promotion activities, etc., to improve user satisfaction and revenue.

[0055] Step S202, obtaining unactivated service data of the target building, generating a decision tree model corresponding to the unactivated service data through the user portrait, and predicting first income data corresponding to the unactivated service data in a future time period through the decision tree model, wherein the decision tree model includes a plurality of sub-nodes, and the unactivated service data indicates a communication service that the user has not activated;

[0056] Specifically, the communication service data of users in the target building that have not yet been activated is obtained. These data may come from market research, industry trend analysis or building expansion strategy planning, including potential service types, expected package charges, circuit levels, etc. The unactivated service data is matched with the user portrait, and the generated user portrait data, especially user characteristics, preferences and usage behaviors, is used to analyze the potential market acceptance of the unactivated service data. The unpruned decision tree strategy in the random forest algorithm is used to build a corresponding decision tree model for each unactivated service data. In the process of building the decision tree, the splitting of each node is based on randomly selected features. The purpose of this is to avoid overfitting and improve the generalization ability of the model. At each node of the decision tree, one or several features are randomly selected from the user portrait for splitting, such as user preferences, usage time period, customer type, etc., to predict the potential income of the unactivated service. The unactivated service data and the corresponding user portrait features are input into the trained decision tree model, so that the decision tree model learns the relationship between different user features and service income, thereby generating a prediction model for the unactivated service data. Based on these features, the model will predict the first revenue data that may be generated by unactivated services in future time periods.

[0057] Step S203, analyze the activated business data through the triple exponential smoothing model, predict the second income data corresponding to the activated business data in the future time period, calculate the sum of the first income data and the second income data, obtain the total income data for the future time period, determine the communication operation plan of the target building according to the total income data and operate it, wherein the communication operation plan represents the planning of communication services for the future time period.

[0058] Specifically, for the types of services that have been opened in commercial buildings, the historical revenue data of the communication services that have been opened in the target buildings are collected, including the monthly revenue and annual revenue of each type of service, which can be further refined to specific service characteristics such as circuit level, package tariff, service type, and rental duration. The above historical revenue data are predicted using the triple exponential smoothing (Holt-Winters) model. The model will analyze the long-term trend, seasonal fluctuations and current level of revenue, and predict the revenue (secondary revenue data) of the services that have been opened in the future time period by adjusting three smoothing parameters (α, β, γ). Because the revenue of the types of services that have been opened in commercial buildings can capture the long-term trend, cyclical fluctuations and current level changes in the time series, it effectively solves the problem that the traditional time series model is not accurate enough when dealing with revenue data with cyclical fluctuations. By training the triple exponential smoothing model, the predicted second revenue data of the services that have been opened in the future time period, such as the next 3 months or longer, are obtained. These data can capture the cyclical and trend changes of revenue based on the model's analysis of historical data.

[0059] Using the user profile and decision tree model generated in the previous step, predict the potential revenue (first revenue data) of unactivated services in the future time period. These predictions are based on user preferences, service types, and market trends, reflecting potential new service growth points. Add the predicted first revenue data (potential revenue of unactivated services) to the second revenue data (predicted revenue of activated services) to obtain the total revenue data for the future time period. Based on the obtained total revenue data and the predicted revenue structure, formulate a communication operation plan. The plan may include: Resource allocation: Based on the predicted revenue growth points, rationally plan network resources and human resources, such as increasing the installation of high-bandwidth circuits and optimizing package structures. Market strategy: Based on the potential revenue of unactivated services, design targeted promotion strategies, such as promotional activities for potential high-value packages. Service optimization: Based on user profiles and revenue forecasts for activated services, optimize customer service and improve customer perception, such as increasing the efficiency of troubleshooting and improving stability.

[0060] Through this embodiment, the activated service data and user information of the target building in the historical time period are obtained, and a user portrait is generated according to the activated service data and user information; the unactivated service data of the target building is obtained, and a decision tree model corresponding to the unactivated service data is generated through the user portrait, and the first income data corresponding to the unactivated service data in the future time period is predicted through the decision tree model; the activated service data is analyzed through the triple exponential smoothing model, and the second income data corresponding to the activated service data in the future time period is predicted, and the sum of the first income data and the second income data is calculated to obtain the total income data of the future time period. Compared with the prior art, when facing complex multi-dimensional business situation characteristics and historical income data, the error of commercial building income prediction is large, resulting in inaccurate corresponding investment and operation plans. In this application, by using different models to predict the activated service data and unactivated service data of the user, the corresponding income can be accurately predicted according to the characteristics of the above data, so as to accurately determine and adjust the subsequent communication operation plan. Therefore, it can solve the problem of large errors in the prediction of communication service costs and inaccurate communication operation planning in the prior art, and achieve the effect of accurately predicting communication service costs and accurately planning operation plans.

[0061] In the specific implementation process, the above step S202 generates a decision tree model corresponding to the unactivated service data through the user portrait, which can be implemented through the following steps: Step S2021: Obtain the user features in the user portrait, and take each data in the unactivated service data as a sub-data set, and modify the sub-data set according to the user features to obtain the modified sub-data set; Step S2022: For each sub-data set, use the unpruned decision tree algorithm to construct a decision tree model, wherein the decision tree model is not pruned during the construction process, and only one feature is selected for judgment when each node is split. Through the above steps, using user portraits and unpruned decision tree algorithms, the potential income of unactivated services in commercial buildings can be effectively predicted, thereby assisting building managers in formulating more accurate business development and income optimization strategies.

[0062] Specifically, based on the user profile formed by the service types that have been opened in the target building, including user preferences, usage time, customer type and other features, random sampling with replacement (Bootstrap Sampling) is adopted. Based on the unpruned decision algorithm, the sub-dataset of the unopened service data is corrected according to the first user feature in the extracted user profile. The correction can be achieved by adjusting the parameters of the algorithm, introducing new feature variables, modifying the decision logic of the algorithm, etc.: Parameter adjustment: adjust the parameters of the algorithm according to the characteristics of the service type, handle missing, abnormal and repeated data in the feature parameters, ensure the quality of the data set, and make it more suitable for specific business scenarios. Feature variable introduction: introduce data such as cost, cycle, package content as new feature variables into the algorithm to improve the accuracy and generalization ability of the algorithm. Decision logic modification: modify or optimize the decision logic of the algorithm according to the business logic and promotion strategy, and adjust the weights of various indicators in the algorithm to make it more in line with actual business scenarios and user needs. Based on each adjusted sub-dataset, a decision tree of a business type is constructed and trained. When constructing a business type decision tree, only one feature is selected for splitting judgment at each node, rather than all features. This strategy prevents the model from overfitting when dealing with complex multi-dimensional features and improves generalization capabilities. It improves the model's inference capabilities and avoids the overfitting problem in traditional decision tree models. In addition, each feature decision tree covers all possible situations. The model improves the accuracy of inference by capturing complex interactive information in the data, especially the interaction between different features.

[0063] In order to improve the overall deduction performance and reduce the deviation between the deduction results and the actual results, the model and strategy are trained as follows: 1. Initial model: Select the algorithm and data, the initial model, denoted as F0(X). 2. Gradually improve the model: At each step m, we use the new model h m (X) to correct the error of the previous step deduction result, and define the constructed decision trees as h m (X). 3. Calculate the residual error: Calculate the income value F derived from the current model m-1 The difference (residual) between (X) and the actual income value is denoted by r i :r i =y i -F m-1 (x i ), where y i is the actual business income value of the i-th sample of a decision tree, X i is the characteristic value of the i-th sample of a decision tree. 4. Fitting a new model: The income deduction residual calculated in the previous step is used to fit a new deduction model h m (X), so that h m(X) Predict the residual as much as possible. 5. Update the model: Add the new model to the existing decision model: F m (X) = F m-1 (X)+γh m (X), where γ is the learning rate, which controls the contribution of each tree to the final model, finds the best feature parameter combination to gradually reduce the difference between the high-income deduction results and the actual income, and improves the accuracy of the deduction results. 6. Repeat steps: Repeat steps 3 to 5 until the model deduction error reaches the preset income error requirement standard. By integrating the deduction results of all feature decision trees, all deduction income results are weighted and summed to obtain the final income deduction. Effectively assist model optimization and improve the accuracy and stability of the income deduction model.

[0064] In some optional implementations, the above step S203 analyzes the opened service data through the triple exponential smoothing model, and predicts the second income data corresponding to the opened service data in the future time period, which can be achieved by the following steps: step S2031: establish a time series index of the opened service data; step S2032: analyze and adjust the opened service data and the corresponding time series index through the triple exponential smoothing model to obtain the smoothing parameters of the triple exponential smoothing model, wherein the smoothing parameters are parameters used to characterize the degree of change of the opened service data over time; step S2033: predict the income of the opened service data in the future time period through the triple exponential smoothing model configured with the smoothing parameters to obtain the second income data. The method uses the triple exponential smoothing model to analyze the income trend of the opened services of the commercial building, and predicts the income in the future time period based on this, that is, the second income data. This prediction method is particularly suitable for processing income data with obvious trends and seasonal fluctuations, and can provide building managers with a clear prediction of future income, so as to formulate corresponding operation strategies and investment plans.

[0065] Specifically, the triple exponential smoothing model is used to deduce the future revenue of the types of services that have been opened in commercial buildings. The advantage of this model is that it can capture the long-term trend, seasonal fluctuations and current level changes in the revenue data at the same time. The specific steps are as follows: According to the historical revenue information in the CRM (Customer Relationship Management, CRM) detailed order data, a time series index is established by month. The historical revenue data of the opened services is analyzed to determine whether there are trends and seasonal components. Trends may be manifested as a gradual increase or decrease in business revenue over time. For example, in some buildings with a large number of e-commerce users, the volume of large-bandwidth broadband services has been increasing year by year; and seasonality may be manifested as a pattern of business revenue repeating in certain time periods (such as quarters and months). Generally, during the Spring Festival, due to the long market closure time, the business volume of commercial buildings before and after the festival often fluctuates greatly. By analyzing the periodicity of historical revenue and the trend of future revenue, the future revenue of the opened business types is deduced. According to the characteristics of business revenue data, we select three smoothing parameters suitable for the situation where the seasonal variation is relatively stable and estimate the model: the average revenue of the services opened in the building is taken as the level parameter α, the historical trend of the revenue of the services opened in the building is taken as the trend parameter β, and the periodic revenue of the services opened in the building is taken as the seasonal parameter γ. Through the training of historical revenue data, the parameters are automatically adjusted to minimize the prediction error to capture the fluctuation and change of the revenue of the same type of business in the same year and period of the building, and finally output the revenue forecast results for 12 months (9 months in the past and 3 months in the future). After the smoothing parameters are determined, the triple exponential smoothing model configured with these parameters can be used to predict the revenue of the services opened in the building in the future time period. The model prediction is based on the analysis of historical revenue data and the adjustment of parameters, and can predict the revenue trend in the next few months or years. For example, to predict the revenue in the next 3 months, the model can obtain the revenue forecast value of each month to form the second revenue data.

[0066] Use historical data or simulated data to verify the revised algorithm and evaluate its performance and accuracy. If problems or deficiencies are found, such as large differences between the deduced results and the actual results, the algorithm can be further optimized and adjusted. Performance evaluation: Evaluate the performance of the algorithm by comparing the predicted results with the actual results, such as accuracy, recall rate, F1 score, etc., to fully measure the performance, correctness and comprehensiveness of the overall deduction of the model. Optimization and adjustment: Optimize and adjust the algorithm as necessary based on the evaluation results, and iterate the optimization process repeatedly to improve its performance and accuracy until the difference between the deduced income and the actual income is within the required error standard. Apply the optimized algorithm to the income deduction of unopened business types in commercial buildings.

[0067] In some optional implementations, after obtaining the activated service data of the target building in the historical time period and obtaining the unactivated service data of the target building, the method further includes the following steps: Step 204: Standardizing the activated service data and the unactivated service data respectively; Step 205: Encoding the activated service data and the unactivated service data after the standardization process to obtain first coded data corresponding to the activated service data and second coded data corresponding to the unactivated service data. Through the above standardization and encoding process, the activated and unactivated service data of the commercial building are converted into a form that can be processed by the model, so that they can be applied to the subsequent modeling and prediction steps, ensuring that the model can more accurately evaluate the potential and trend of business income and provide more powerful data support for the operation and management of the building.

[0068] Table 1 Mapping table of characteristic data and numerical codes

[0069]

[0070]

[0071] Specifically, the revenue characteristics of the opened business types and the opened business types of commercial buildings are standardized to ensure that the features of different magnitudes have the same scale; the corresponding circuit level, package tariff, business type, rental duration, and customer perception category features are converted into numerical codes to ensure that the classification variables can effectively participate in model training. Table 1 is a mapping table of feature data and numerical codes, as shown in Table 1.

[0072] In some other optional implementations, the above step S201 generates a user portrait based on the activated service data and user information, including step S2011: extracting key features that have a significant impact on user behavior from the activated service data; step S2012: dividing users into different groups by statistical methods and analyzing the behavior patterns of each group; step S2013: constructing a user portrait by combining key features and behavior patterns. This method can generate detailed user portraits that not only reflect the characteristics of individual users, but also reveal the behavior patterns of groups, which is crucial for predicting potential revenue of different types of services, optimizing service and package design, formulating market strategies, and improving customer satisfaction.

[0073] Specifically, first, extract those features that have a significant impact on user behavior from the activated service data. This may include but is not limited to: Service type: What type of communication service does the user prefer, such as broadband, fixed-line, government and enterprise dedicated line, etc. Package selection: The package level or tariff selected by the user, which reflects the user's sensitivity to service quality and price. Frequency and time of use: The user's usage habits during the day or week, such as peak usage, night usage, or differences between weekdays and weekends. Rental duration: The duration of the user's rental service, which helps to understand user loyalty and potential renewal behavior. Customer perception: User satisfaction and experience reflected by data such as complaints, reviews, or surveys. Payment behavior: User payment method preferences, such as automatic deduction, installment payment, or one-time payment. These features need to be ensured through data preprocessing and feature engineering, such as data cleaning, missing value processing, outlier detection, etc. to ensure their quality and availability. Use statistical methods and clustering algorithms, such as K-means clustering, hierarchical clustering, DBSCAN, etc. to divide users into different groups. The basis for grouping is the key characteristics of users and their behavior patterns. Analyze the behavioral patterns of each group, which may involve: Differences in usage patterns, such as the difference between users with high bandwidth requirements and users with low bandwidth requirements. Consumption levels and preferences, identifying high-value customers and potential churn customers, etc. User satisfaction and feedback, understanding which services or packages have received positive reviews from users. User renewal and upgrade behavior, predicting the user's possible future behavior. Build user portraits based on the analysis of key features and behavioral patterns. User portraits may include: Basic information about users, such as age, gender, occupation, etc. User preferences and interests, such as preferences for package types and service time. User consumption habits, such as average monthly consumption, payment method preferences, etc. User satisfaction and loyalty, based on customer perception and renewal behavior. User potential needs, predictions of future service needs based on behavioral pattern analysis. The construction of user portraits requires comprehensive consideration of the degree of influence of various features on user behavior, and may use techniques such as feature weight adjustment, feature selection, or feature combination to ensure the accuracy and practicality of the portraits.

[0074] In some optional implementations, after obtaining the total revenue data for the future time period, the method further includes step S206: obtaining actual revenue data; step S207: calculating the error parameter between the actual revenue data and the total revenue data to evaluate the prediction accuracy, wherein the error parameter is a parameter that characterizes the error size. By calculating the error parameter, the method provides a quantitative evaluation standard for the model, helping building managers understand the limitations of the model prediction and take into account the uncertainty of the prediction when making decisions.

[0075] Specifically, ensure that actual revenue data for commercial buildings in the future time period is obtained. This data is usually collected after the end of the predicted future time period through the building's financial system, CRM system, or other business management system. The actual revenue data needs to be aligned in time with the predicted revenue data to enable accurate error calculations. Calculate error parameters: Next, use different error metrics to evaluate the accuracy of the forecast. Common error parameters include: Mean Squared Error (MSE): Calculate the average of the squared differences between the predicted value and the actual value. The smaller the MSE, the more accurate the forecast. Root Mean Squared Error (RMSE): The square root of MSE. RMSE is a variant of MSE and more intuitively reflects the size of the forecast error. Mean Absolute Error (MAE): Calculate the average of the absolute difference between the predicted value and the actual value. MAE represents the average size of the forecast error. There is no square effect, so it is more sensitive to larger errors. Coefficient of Determination is also known as R 2 , R 2 The value measures the model's ability to explain the data variance, ranging from 0 to 1. 2 The closer it is to 1, the better the model fits the data. Based on the calculated error parameters, analyze the accuracy of the forecast results. If the error parameters show a significant deviation between the forecast and the actual revenue, the model needs to be adjusted, which may include feature selection, model parameter optimization, algorithm adjustment, etc. to improve the accuracy of the forecast. This step is crucial to the continuous improvement of the forecast model. Through continuous evaluation and improvement, the model can be ensured to be more reliable and effective in future forecasts.

[0076] In some optional implementations, the above step S201 obtains the activated service data and user information of the target building in the historical time period, including step S2014: obtaining the user information and the activated service data of the user by matching the service number, wherein the service number is a unique number generated by the user in the process of activating the service. This method can accurately obtain the historical activated service data and user information of all users in the target building, providing a solid data foundation for subsequent revenue deduction and user behavior analysis. The use of service numbers ensures the accuracy and efficiency of data matching, while data integration and preprocessing ensure the reliability and effectiveness of the analysis results.

[0077] Specifically, use the service number as the query key to perform precise matching in the database. Extract user information from the CRM system, such as the user's basic information (name, contact information, service activation time, etc.) and user feedback information (such as satisfaction surveys, service complaint records, etc.). Obtain the activated service data from the service order system, such as circuit level, package tariff, service type, rental duration, etc. Ensure that the above data corresponds to the service number one by one to ensure the accuracy and completeness of the acquired data. Integrate the matched user information and activated service data together to form a comprehensive data set containing user characteristics and historical service consumption records. This may include: Service revenue: the fees paid by users for activated services during the historical time period. Service usage: the frequency, usage duration, peak usage time, etc. of users' use of activated services. User satisfaction: the user's evaluation of activated services, which can be collected through questionnaires, online ratings, etc. User classification: users are divided into different groups based on their usage behavior and consumption habits, such as high-value users, ordinary users, and potential churn users.

[0078] In order to enable those skilled in the art to more clearly understand the technical solution of the present application, the implementation process of the method for predicting communication service fees of the present application will be described in detail below in conjunction with specific embodiments.

[0079] This embodiment relates to a specific method for predicting communication service fees. Figure 3 As shown, the following steps are included:

[0080] Step S1: Collecting business income information and user information of commercial buildings within a preset time period;

[0081] Step S2: Create a data set based on the basic information of the services that have been opened in the commercial building (including circuit level, package tariff, service type, rental duration, customer perception, etc.) and the portrait features formed by the user information of the services that have been opened in the building;

[0082] Step S3: Select corresponding characteristic parameters for the two types of commercial building incomes, deduce future incomes based on the algorithm and model, and train the model to obtain the first deduction result;

[0083] Step S4: Evaluate the model, analyze and calculate the residuals between the deduction results and the actual results, fit the new model, adjust and optimize the original model, and continue to train the commercial building business income model until the model deduction error reaches the preset income error requirement standard;

[0084] Step S5: Integrate the income deduction results of the two types of commercial buildings, and finally output the income deduction results for 12 months (9 months in the past and the next three months) after weighting.

[0085] This embodiment also relates to a specific communication service fee prediction and deduction main line logic diagram, such as Figure 4 As shown, the following steps are included:

[0086] Step S6: Known business data is input into the business model through the model, and the model is calculated to obtain unknown possible results, and deduction support is provided to obtain strategic decisions;

[0087] Step S7: The above steps are divided into model backend calculation (model capability construction process, corresponding to prediction and deduction) and model frontend application (model capability reuse process, corresponding to application center);

[0088] Step S8: Known business data belongs to the business parameters of the model front end, and the data is mapped to the known business data of the model back end. The model is fed to the existing business model (a group or one) at the model back end for prediction and deduction, and further to the model back end model deduction results (a group or one), and the deduction results are mapped to the model front end (business deduction results), including 1. report statistics, 2. icon display, and 3. decision recommendations.

[0089] Figure 5 A schematic diagram of a decision tree legend is shown, and the head node of the decision tree includes activation (service data has been activated): 9, not activation (service data has not been activated): 5; then it is divided into three nodes according to user preferences: office: activation 4, not activation 1; daily Internet access: activation 2, not activation 3; live broadcast, game: activation 3, not activation 1; office traffic: faster: activation 4, not activation 0; general: activation 0, not activation 1; daily Internet access charges: higher, activation 0, not activation 2; average level: activation 2, not activation 1; live broadcast, game stability, better: activation 3, not activation 0, general: activation 0, not activation 1.

[0090] Figure 6 A flow chart of a deduction model is shown, which collects historical data on building business income in the past 2-3 years; analyzes the building business income data to determine whether there are trends and seasonal components; the monthly income of some business types is flat, and the seasonal change range is relatively stable; in this case, the additive model is selected and the parameters are evaluated; the horizontal parameter α: the average income of the services already opened in the building; the trend parameter β: the historical trend of the income of the services already opened in the building; the seasonal parameter γ: the cyclical income of the services already opened in the building, and the model is trained based on the historical data of commercial buildings and the selected parameters; outputs the income forecast results for 12 months (9 for the next 3 months); evaluates the difference between the income forecast results and the actual income, and adjusts and optimizes the name based on the difference results.

[0091] Figure 7A line chart of building income deduction is shown, with the horizontal axis being the month and the vertical axis being the income. Based on the actual income from January 2021 to December 2023, the income from January to December 2024 is predicted, and the actual income from January to September 2024 is also obtained and compared with the predicted income. Figure 7 It can be seen that the trends are consistent, indicating that the predicted income obtained through this application is relatively accurate.

[0092] Figure 8 A flowchart combining the Holt-Winters triple exponential smoothing model and the random forest algorithm is shown, including data reading, data clipping, feature processing, and whether the business type is "accepted". If so, the Holt-Winters triple exponential smoothing model is used to predict the income. If not, the random forest algorithm is used to predict the income. The model is then evaluated, and finally the results are integrated and the report is generated.

[0093] The embodiment of the present application also provides a prediction device for communication service costs. It should be noted that the prediction device for communication service costs of the embodiment of the present application can be used to execute the prediction method for communication service costs provided by the embodiment of the present application. The device is used to implement the above-mentioned embodiments and preferred implementation modes, and the descriptions have been made and will not be repeated. As used below, the term "module" can implement a combination of software and / or hardware for a predetermined function. Although the device described in the following embodiments is preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and conceived.

[0094] The following is an introduction to the communication service fee prediction device provided in an embodiment of the present application.

[0095] Fig. 9 Schematic diagram of a communication service fee prediction device according to an embodiment of the present application. Fig. 9 As shown, the device comprises:

[0096] The generating unit 10 is used to obtain the activated service data and user information of the target building in the historical time period, and generate a user portrait according to the activated service data and the user information, wherein the activated service data indicates the communication service that the user has activated, the user information indicates the identity and characteristics of the user, and the user portrait is a virtual image representing the characteristics of the user;

[0097] Specifically, this real-time example uses buildings as units to count the communication costs of a building. The building being counted is the target building, and the activated service data and corresponding user information of users in the target building in the historical time period are extracted from the service order data. The activated service data includes but is not limited to: service type, package tariff, circuit level, rental duration, etc. At the same time, the collected user information, such as customer type, usage habits, preferences, etc., can be obtained from user registration information, market research or user feedback. Based on the above information, a user portrait of the user can be constructed. The user portrait includes a variety of user portrait features, including but not limited to user preferences, usage time period, customer type, and device type readiness information. The above user portrait features are obtained by in-depth analysis and summary of the above activated service data and user information, which not only characterizes the identity and characteristics of the user, but also reflects the user's actual demand and potential value for the building communication service. The generation and optimization of user portraits can not only help predict the communication service revenue of the building, but also guide the market strategy and operation decision of the building, such as optimizing services, adjusting packages, increasing promotion activities, etc., to improve user satisfaction and revenue.

[0098] The prediction unit 20 is used to obtain the unactivated service data of the target building, generate a decision tree model corresponding to the unactivated service data through the user portrait, and predict the first income data corresponding to the unactivated service data in the future time period through the decision tree model, wherein the decision tree model includes a plurality of sub-nodes, and the unactivated service data indicates the communication service that the user has not activated;

[0099] Specifically, the communication service data of users in the target building that have not yet been activated is obtained. These data may come from market research, industry trend analysis or building expansion strategy planning, including potential service types, expected package charges, circuit levels, etc. The unactivated service data is matched with the user portrait, and the generated user portrait data, especially user characteristics, preferences and usage behaviors, is used to analyze the potential market acceptance of the unactivated service data. The unpruned decision tree strategy in the random forest algorithm is used to build a corresponding decision tree model for each unactivated service data. In the process of building the decision tree, the splitting of each node is based on randomly selected features. The purpose of this is to avoid overfitting and improve the generalization ability of the model. At each node of the decision tree, one or several features are randomly selected from the user portrait for splitting, such as user preferences, usage time period, customer type, etc., to predict the potential income of the unactivated service. The unactivated service data and the corresponding user portrait features are input into the trained decision tree model, so that the decision tree model learns the relationship between different user features and service income, thereby generating a prediction model for the unactivated service data. Based on these features, the model will predict the first revenue data that may be generated by unactivated services in future time periods.

[0100] The calculation unit 30 is used to analyze the activated business data through a triple exponential smoothing model, predict the second income data corresponding to the activated business data in the future time period, calculate the sum of the first income data and the second income data, obtain the total income data in the future time period, determine the communication operation plan of the target building according to the total income data and operate it, wherein the communication operation plan represents the planning of communication services in the future time period.

[0101] Specifically, for the types of services that have been opened in commercial buildings, the historical revenue data of the communication services that have been opened in the target buildings are collected, including the monthly revenue and annual revenue of each type of service, which can be further refined to specific service characteristics such as circuit level, package tariff, service type, and rental duration. The above historical revenue data are predicted using the triple exponential smoothing (Holt-Winters) model. The model will analyze the long-term trend, seasonal fluctuations and current level of revenue, and predict the revenue (secondary revenue data) of the services that have been opened in the future time period by adjusting three smoothing parameters (α, β, γ). Because the revenue of the types of services that have been opened in commercial buildings can capture the long-term trend, cyclical fluctuations and current level changes in the time series, it effectively solves the problem that the traditional time series model is not accurate enough when dealing with revenue data with cyclical fluctuations. By training the triple exponential smoothing model, the predicted second revenue data of the services that have been opened in the future time period, such as the next 3 months or longer, are obtained. These data can capture the cyclical and trend changes of revenue based on the model's analysis of historical data.

[0102] Through this embodiment, the activated service data and user information of the target building in the historical time period are obtained, and a user portrait is generated according to the activated service data and user information; the unactivated service data of the target building is obtained, and a decision tree model corresponding to the unactivated service data is generated through the user portrait, and the first income data corresponding to the unactivated service data in the future time period is predicted through the decision tree model; the activated service data is analyzed through the triple exponential smoothing model, and the second income data corresponding to the activated service data in the future time period is predicted, and the sum of the first income data and the second income data is calculated to obtain the total income data of the future time period. Compared with the prior art, when facing complex multi-dimensional business situation characteristics and historical income data, the error of commercial building income prediction is large, resulting in inaccurate corresponding investment and operation plans. In this application, by using different models to predict the activated service data and unactivated service data of the user, the corresponding income can be accurately predicted according to the characteristics of the above data, so as to accurately determine and adjust the subsequent communication operation plan. Therefore, it can solve the problem of large errors in the prediction of communication service costs and inaccurate communication operation planning in the prior art, and achieve the effect of accurately predicting communication service costs and accurately planning operation plans.

[0103] In the specific implementation process, the prediction unit includes a correction module and a first construction module. The correction module is used to obtain user features in the user portrait, and each data in the unactivated service data is used as a sub-data set, and the sub-data set is corrected according to the user features to obtain the corrected sub-data set; the first construction module is used to construct a decision tree model for each sub-data set using an unpruned decision tree algorithm, wherein the decision tree model is not pruned during the construction process, and only one feature is selected for judgment when each node is split. Through the above steps, using user portraits and unpruned decision tree algorithms, the potential income of unactivated services in commercial buildings can be effectively predicted, thereby assisting building managers in formulating more accurate business development and income optimization strategies.

[0104] Specifically, based on the user profile formed by the service types that have been opened in the target building, including user preferences, usage time, customer type and other features, random sampling with replacement (Bootstrap Sampling) is adopted. Based on the unpruned decision algorithm, the sub-dataset of the unopened service data is corrected according to the first user feature in the extracted user profile. The correction can be achieved by adjusting the parameters of the algorithm, introducing new feature variables, modifying the decision logic of the algorithm, etc.: Parameter adjustment: adjust the parameters of the algorithm according to the characteristics of the service type, handle missing, abnormal and repeated data in the feature parameters, ensure the quality of the data set, and make it more suitable for specific business scenarios. Feature variable introduction: introduce data such as cost, cycle, package content as new feature variables into the algorithm to improve the accuracy and generalization ability of the algorithm. Decision logic modification: modify or optimize the decision logic of the algorithm according to the business logic and promotion strategy, and adjust the weights of various indicators in the algorithm to make it more in line with actual business scenarios and user needs. Based on each adjusted sub-dataset, a decision tree of a business type is constructed and trained. When constructing a business type decision tree, only one feature is selected for splitting judgment at each node, rather than all features. This strategy prevents the model from overfitting when dealing with complex multi-dimensional features and improves generalization capabilities. It improves the model's inference capabilities and avoids the overfitting problem in traditional decision tree models. In addition, each feature decision tree covers all possible situations. The model improves the accuracy of inference by capturing complex interactive information in the data, especially the interaction between different features.

[0105] In order to improve the overall deduction performance and reduce the deviation between the deduction results and the actual results, the model and strategy are trained as follows: 1. Initial model: Select the algorithm and data, the initial model, denoted as F0(X). 2. Gradually improve the model: At each step m, we use the new model h m (X) to correct the error of the previous step deduction result, and define the constructed decision trees as hm (X). 3. Calculate the residual error: Calculate the income value F derived from the current model m-1 The difference (residual) between (X) and the actual income value is denoted by r i :r i =y i -F m-1 (x i ), where y i is the actual business income value of the i-th sample of a decision tree, X i is the characteristic value of the i-th sample of a decision tree. 4. Fitting a new model: The income deduction residual calculated in the previous step is used to fit a new deduction model h m (X), so that h m (X) Predict the residual as much as possible. 5. Update the model: Add the new model to the existing decision model: F m (X) = F m-1 (X)+γh m (X), where γ is the learning rate, which controls the contribution of each tree to the final model, finds the best feature parameter combination to gradually reduce the difference between the high-income deduction results and the actual income, and improves the accuracy of the deduction results. 6. Repeat steps: Repeat steps 3 to 5 until the model deduction error reaches the preset income error requirement standard. By integrating the deduction results of all feature decision trees, all deduction income results are weighted and summed to obtain the final income deduction. Effectively assist model optimization and improve the accuracy and stability of the income deduction model.

[0106] In some optional embodiments, the above-mentioned calculation unit includes an establishment module, an adjustment module and a configuration module, the establishment module is used to establish a time series index of the opened service data; the adjustment module is used to analyze and adjust the opened service data and the corresponding time series index through a triple exponential smoothing model to obtain the smoothing parameters of the triple exponential smoothing model, wherein the smoothing parameters are parameters used to characterize the degree of change of the opened service data over time; the configuration module is used to predict the income of the opened service data in the future time period through the triple exponential smoothing model configured with the smoothing parameters to obtain the second income data. The device uses the triple exponential smoothing model to analyze the income trend of the opened services of the commercial building, and based on this, predicts the income in the future time period, i.e., the second income data. This prediction device is particularly suitable for processing income data with obvious trends and seasonal fluctuations, and can provide building managers with a clear forecast of future income, so as to formulate corresponding operation strategies and investment plans.

[0107] Specifically, the triple exponential smoothing model is used to deduce the future revenue of the types of services that have been opened in commercial buildings. The advantage of this model is that it can capture the long-term trend, seasonal fluctuations and current level changes in the revenue data at the same time. The specific steps are as follows: According to the historical revenue information in the CRM (Customer Relationship Management, CRM) detailed order data, a time series index is established by month. The historical revenue data of the opened services is analyzed to determine whether there are trends and seasonal components. Trends may be manifested as a gradual increase or decrease in business revenue over time. For example, in some buildings with a large number of e-commerce users, the volume of large-bandwidth broadband services has been increasing year by year; and seasonality may be manifested as a pattern of business revenue repeating in certain time periods (such as quarters and months). Generally, during the Spring Festival, due to the long market closure time, the business volume of commercial buildings before and after the festival often fluctuates greatly. By analyzing the periodicity of historical revenue and the trend of future revenue, the future revenue of the opened business types is deduced. According to the characteristics of business revenue data, we select three smoothing parameters suitable for the situation where the seasonal variation is relatively stable and estimate the model: the average revenue of the services opened in the building is taken as the level parameter α, the historical trend of the revenue of the services opened in the building is taken as the trend parameter β, and the periodic revenue of the services opened in the building is taken as the seasonal parameter γ. Through the training of historical revenue data, the parameters are automatically adjusted to minimize the prediction error to capture the fluctuation and change of the revenue of the same type of business in the same year and period of the building, and finally output the revenue forecast results for 12 months (9 months in the past and 3 months in the future). After the smoothing parameters are determined, the triple exponential smoothing model configured with these parameters can be used to predict the revenue of the services opened in the building in the future time period. The model prediction is based on the analysis of historical revenue data and the adjustment of parameters, and can predict the revenue trend in the next few months or years. For example, to predict the revenue in the next 3 months, the model can obtain the revenue forecast value of each month to form the second revenue data.

[0108] In some optional implementations, after obtaining the activated service data of the target building in the historical time period and obtaining the unactivated service data of the target building, the device further includes a standardization processing unit and an encoding unit, the standardization processing unit is used to perform standardization processing on the activated service data and the unactivated service data respectively; the encoding unit is used to encode the activated service data and the unactivated service data after the standardization processing to obtain the first encoded data corresponding to the activated service data and the second encoded data corresponding to the unactivated service data. Through the above-mentioned standardization and encoding processing, the activated and unactivated service data of the commercial building are converted into a form that can be processed by the model, so that it can be applied to the subsequent modeling and prediction steps, ensuring that the model can more accurately evaluate the potential and trend of business income and provide more powerful data support for the operation and management of the building.

[0109] Specifically, the revenue characteristics of the business types that have been opened and the characteristics of the business types that have been opened in commercial buildings are standardized to ensure that features of different magnitudes have the same scale; the corresponding circuit levels, package rates, business types, rental durations, and customer perception category characteristics are converted into numerical codes to ensure that categorical variables can effectively participate in model training.

[0110] In some other optional implementations, the generation unit includes an extraction module, an analysis module, and a second construction module. The extraction module is used to extract key features that have a significant impact on user behavior from the activated service data; the analysis module is used to divide users into different groups through a statistical device and analyze the behavior pattern of each group; the second construction module is used to construct a user portrait by combining key features and behavior patterns. The device can generate detailed user portraits that not only reflect the characteristics of individual users, but also reveal the behavior patterns of groups, which is crucial for predicting potential revenue of different types of services, optimizing service and package design, formulating market strategies, and improving customer satisfaction.

[0111] Specifically, first, extract those features that have a significant impact on user behavior from the data of the activated services. This may include but is not limited to: Service type: What type of communication service does the user prefer, such as broadband, fixed-line, government and enterprise dedicated line, etc. Package selection: The package level or tariff selected by the user, which reflects the user's sensitivity to service quality and price. Frequency and time of use: The user's usage habits during the day or week, such as peak usage, night usage, or differences between weekdays and weekends. Rental duration: The duration of the user's rental service, which helps to understand user loyalty and potential renewal behavior. Customer perception: User satisfaction and experience reflected by data such as complaints, reviews, or surveys. Payment behavior: User payment method preferences, such as automatic deduction, installment payment, or one-time payment. These features need to be ensured through data preprocessing and feature engineering, such as data cleaning, missing value processing, outlier detection, etc. to ensure their quality and availability. Use statistical devices and clustering algorithms, such as K-means clustering, hierarchical clustering, DBSCAN, etc. to divide users into different groups. The basis for grouping is the key characteristics of users and their behavior patterns. Analyze the behavioral patterns of each group, which may involve: Differences in usage patterns, such as the difference between users with high bandwidth requirements and users with low bandwidth requirements. Consumption levels and preferences, identifying high-value customers and potential churn customers, etc. User satisfaction and feedback, understanding which services or packages have received positive reviews from users. User renewal and upgrade behavior, predicting the user's possible future behavior. Build user portraits based on the analysis of key features and behavioral patterns. User portraits may include: Basic information about users, such as age, gender, occupation, etc. User preferences and interests, such as preferences for package types and service time. User consumption habits, such as average monthly consumption, payment method preferences, etc. User satisfaction and loyalty, based on customer perception and renewal behavior. User potential needs, predictions of future service needs based on behavioral pattern analysis. The construction of user portraits requires comprehensive consideration of the degree of influence of various features on user behavior, and may use techniques such as feature weight adjustment, feature selection, or feature combination to ensure the accuracy and practicality of the portraits.

[0112] In some optional embodiments, after obtaining the total income data for the future time period, the device further includes an acquisition unit and a calculation unit, wherein the acquisition unit is used to acquire the actual income data; and the calculation unit is used to calculate the error parameter between the actual income data and the total income data to evaluate the prediction accuracy, wherein the error parameter is a parameter that characterizes the error size. By calculating the error parameter, the device provides a quantitative evaluation standard for the model, helping building managers understand the limitations of the model prediction and take into account the uncertainty of the prediction when making decisions.

[0113] Specifically, ensure that actual revenue data for commercial buildings in the future time period is obtained. This data is usually collected after the end of the predicted future time period through the building's financial system, CRM system, or other business management system. The actual revenue data needs to be aligned in time with the predicted revenue data to enable accurate error calculations. Calculate error parameters: Next, use different error metrics to evaluate the accuracy of the forecast. Common error parameters include: Mean Squared Error (MSE): Calculate the average of the squared differences between the predicted value and the actual value. The smaller the MSE, the more accurate the forecast. Root Mean Squared Error (RMSE): The square root of MSE. RMSE is a variant of MSE and more intuitively reflects the size of the forecast error. Mean Absolute Error (MAE): Calculate the average of the absolute difference between the predicted value and the actual value. MAE represents the average size of the forecast error. There is no square effect, so it is more sensitive to larger errors. Coefficient of Determination is also known as R 2 , R 2 The value measures the model's ability to explain the data variance, ranging from 0 to 1. 2 The closer it is to 1, the better the model fits the data. Based on the calculated error parameters, analyze the accuracy of the forecast results. If the error parameters show a significant deviation between the forecast and the actual revenue, the model needs to be adjusted, which may include feature selection, model parameter optimization, algorithm adjustment, etc. to improve the accuracy of the forecast. This step is crucial to the continuous improvement of the forecast model. Through continuous evaluation and improvement, the model can be ensured to be more reliable and effective in future forecasts.

[0114] In some optional implementations, the generation unit includes a matching module for obtaining user information and user activated service data through service number matching, wherein the service number is a unique number generated by the user in the process of activating the service. The device can accurately obtain the historical activated service data and user information of all users in the target building, providing a solid data foundation for subsequent revenue deduction and user behavior analysis. The use of service numbers ensures the accuracy and efficiency of data matching, while data integration and preprocessing ensure the reliability and effectiveness of the analysis results.

[0115] Specifically, use the service number as the query key to perform precise matching in the database. Extract user information from the CRM system, such as the user's basic information (name, contact information, service activation time, etc.) and user feedback information (such as satisfaction surveys, service complaint records, etc.). Obtain the activated service data from the service order system, such as circuit level, package tariff, service type, rental duration, etc. Ensure that the above data corresponds to the service number one by one to ensure the accuracy and completeness of the acquired data. Integrate the matched user information and activated service data together to form a comprehensive data set containing user characteristics and historical service consumption records. This may include: Service revenue: the fees paid by users for activated services during the historical time period. Service usage: the frequency, usage duration, peak usage time, etc. of users' use of activated services. User satisfaction: the user's evaluation of activated services, which can be collected through questionnaires, online ratings, etc. User classification: users are divided into different groups based on their usage behavior and consumption habits, such as high-value users, ordinary users, and potential churn users.

[0116] The communication service fee prediction device includes a processor and a memory. The generation unit, prediction unit, and calculation unit are stored in the memory as program units, and the processor executes the program units stored in the memory to implement corresponding functions. The modules are all located in the same processor; or, the modules are located in different processors in any combination.

[0117] The processor includes a kernel, which retrieves the corresponding program unit from the memory. One or more kernels can be set, and the kernel parameters can be adjusted to solve the problem of large errors in the forecast of communication income and expenses of commercial buildings, resulting in inaccurate corresponding investment and operation plans.

[0118] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM, and the memory includes at least one memory chip.

[0119] An embodiment of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium includes a stored program, wherein when the program is executed, a device where the computer-readable storage medium is located is controlled to execute a method for predicting communication service fees.

[0120] An embodiment of the present invention provides an electronic device, including a processor, a memory, and a program stored in the memory and executable on the processor. When the processor executes the program, at least the above steps are implemented.

[0121] The devices in this article can be servers, PCs, PADs, mobile phones, etc.

[0122] The present application also provides a computer program product, including a computer program, which implements the methods in various embodiments of the present application when executed by a processor.

[0123] Obviously, those skilled in the art should understand that the above modules or steps of the present invention can be implemented by a general computing device, they can be concentrated on a single computing device, or distributed on a network composed of multiple computing devices, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, and in some cases, the steps shown or described can be executed in a different order than here, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. Thus, the present invention is not limited to any specific combination of hardware and software.

[0124] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0125] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0126] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0127] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0128] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0129] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0130] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0131] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0132] From the above description, it can be seen that the above embodiments of the present application achieve the following technical effects:

[0133] 1) In the communication service fee prediction method of the present application, the activated service data and user information of the target building in the historical time period are obtained, and a user portrait is generated based on the activated service data and user information; the unactivated service data of the target building is obtained, and a decision tree model corresponding to the unactivated service data is generated through the user portrait, and the first income data corresponding to the unactivated service data in the future time period is predicted through the decision tree model; the activated service data is analyzed through the triple exponential smoothing model to predict the second income data corresponding to the activated service data in the future time period, and the sum of the first income data and the second income data is calculated to obtain the total income data for the future time period. Compared with the prior art, when faced with complex multi-dimensional business situation characteristics and historical income data, the error of commercial building income prediction is large, resulting in inaccurate corresponding investment and operation plans. In this application, by using different models to predict the user's activated service data and unactivated service data, the corresponding income can be accurately predicted according to the characteristics of the above data, so as to accurately determine and adjust the subsequent communication operation plan. Therefore, it is possible to solve the problems of large errors in predicting communication service costs and inaccurate communication operation planning in the prior art, and achieve the effect of accurately predicting communication service costs and accurately planning operation plans.

[0134] The above description is only the preferred embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A method for predicting communication service costs, characterized in that: include: Acquire the activated service data and user information of the target building in the historical time period, and generate a user portrait according to the activated service data and the user information, wherein the activated service data indicates the communication service that the user has activated, the user information indicates the identity and characteristics of the user, and the user portrait is a virtual image that at least represents the characteristics of the user; Acquire the unactivated service data of the target building, generate a decision tree model corresponding to the unactivated service data through the user portrait, and predict the first income data corresponding to the unactivated service data in a future time period through the decision tree model, wherein the decision tree model includes a plurality of sub-nodes, and the unactivated service data represents the communication service that the user has not activated; The activated business data is analyzed by a triple exponential smoothing model to predict second revenue data corresponding to the activated business data in the future time period, and the sum of the first revenue data and the second revenue data is calculated to obtain total revenue data for the future time period. The communication operation plan of the target building is determined based on the total revenue data and the operation is performed, wherein the communication operation plan represents the planning of communication services for the future time period.

2. The method for predicting communication service fees according to claim 1, characterized in that: Generating a decision tree model corresponding to the unactivated service data through the user portrait includes: Acquire user features in the user portrait, and use each data in the unactivated service data as a sub-data set, and modify the sub-data set according to the user features to obtain the modified sub-data set; For each of the sub-data sets, the decision tree model is constructed using an unpruned decision tree algorithm, wherein the decision tree model is not pruned during the construction process, and only one feature is selected for judgment when each node is split.

3. The method for predicting communication service fees according to claim 1, characterized in that: Analyzing the activated service data by a triple exponential smoothing model to predict and obtain second income data corresponding to the activated service data in the future time period includes: Establishing a time series index of the activated service data; Analyzing and adjusting the activated service data and the corresponding time series index through a triple exponential smoothing model to obtain a smoothing parameter of the triple exponential smoothing model, wherein the smoothing parameter is a parameter used to characterize the degree of change of the activated service data over time; The revenue of the activated service data in the future time period is predicted by a triple exponential smoothing model configured with the smoothing parameters to obtain the second revenue data.

4. The method for predicting communication service fees according to claim 1, characterized in that: After obtaining the activated service data of the target building in the historical time period and obtaining the unactivated service data of the target building, the method further includes: Performing standardization processing on the activated service data and the unactivated service data respectively; The activated service data and the unactivated service data after the standardization process are encoded to obtain first encoded data corresponding to the activated service data and second encoded data corresponding to the unactivated service data.

5. The method for predicting communication service fees according to claim 1, characterized in that: Generating a user profile according to the activated service data and the user information includes: Extracting key features that have a significant impact on user behavior from the activated service data; Dividing the users into different groups by statistical methods and analyzing the behavior patterns of each group; The user portrait is constructed by combining the key features and the behavior patterns.

6. The method for predicting communication service fees according to claim 1, characterized in that: After obtaining the total revenue data for the future time period, the method further includes: Obtain actual income data; The error parameter between the actual revenue data and the total revenue data is calculated to evaluate the prediction accuracy, wherein the error parameter is a parameter that characterizes the size of the error.

7. The method for predicting communication service fees according to claim 1, characterized in that: Obtain the target building's service data and user information for the historical period, including: The user information and the activated service data of the user are obtained by matching the service number, wherein the service number is a unique number generated by the user in the process of activating the service.

8. A communication service fee prediction device, characterized in that: include: A generating unit, configured to obtain activated service data and user information of a target building in a historical time period, and generate a user portrait according to the activated service data and the user information, wherein the activated service data indicates the communication service activated by the user, the user information indicates the identity and characteristics of the user, and the user portrait is a virtual image representing the characteristics of the user; a prediction unit, configured to obtain unactivated service data of the target building, generate a decision tree model corresponding to the unactivated service data through the user portrait, and predict first income data corresponding to the unactivated service data in a future time period through the decision tree model, wherein the decision tree model includes a plurality of sub-nodes, and the unactivated service data represents a communication service that the user has not activated; A calculation unit is used to analyze the activated business data through a triple exponential smoothing model, predict the second income data corresponding to the activated business data in the future time period, calculate the sum of the first income data and the second income data, obtain the total income data for the future time period, and determine the communication operation plan of the target building according to the total income data and perform operations, wherein the communication operation plan represents the planning of communication services for the future time period.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored program, wherein when the program is executed, the device where the computer-readable storage medium is located is controlled to execute the communication service fee prediction method according to any one of claims 1 to 7.

10. An electronic device, characterized in that: include: One or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and are configured to be executed by the one or more processors, and the one or more programs include a method for predicting communication service costs according to any one of claims 1 to 7.

Images