Method and system for predicting total active days per user on basis of stepwise supervised learning

Abstract

Disclosed in the present application are a method and system for predicting total active days (TAD) per user on the basis of stepwise supervised learning. The method comprises: grouping channel data within a set time window by month, and extracting a multi-dimensional feature set; performing initial training and optimization by using a supervised machine learning model; progressively expanding training samples by means of an iterative prediction mechanism, and performing stage-wise prediction of TAD values over shortened periods until all stages are covered; and finally, constructing a prediction model on the basis of a complete dataset, so as to achieve accurate prediction of newly added channel data. By means of stepwise prediction, the present invention maximizes the utilization of recent data as samples, thereby more accurately reflecting the current status of a product and user activity behavior patterns, and thus significantly improving the prediction accuracy. In addition, the stepwise prediction strategy enables the model to gradually adapt to evolving data trends, and still maintain relatively high prediction performance under frequent product iterations and significant changes in channel structures.

Description

A Method and System for Predicting Total User Activity Days Based on Step-by-Step Supervised Learning Technical Field

[0001] This invention relates to the field of machine learning technology, and in particular to a method and system for predicting total active days of users based on step-by-step supervised learning. Background Technology

[0002] In the current internet and mobile internet landscape, user behavior analysis is crucial for product optimization, marketing, and user experience improvement. Total Active Days per User (TAD) is a key metric for measuring user stickiness and activity. It refers to the total number of days each user is active, i.e., the average number of days each user uses a particular application within a given statistical period. If the statistical period is longer than one year, TAD can essentially reflect the number of days a user used the application before churning. For operations teams and product developers, accurately predicting TAD helps in developing more effective operational strategies, optimizing product features, and enhancing user experience.

[0003] However, traditional supervised TAD prediction methods face numerous challenges in practical applications. Traditional methods typically require historical data over a long period, such as data from 360 days ago, for model training and prediction. However, this approach has several significant problems. First, over time, the product itself may have undergone multiple iterations, with significant changes in functionality and interface, leading to alterations in user behavior patterns and activity levels. Therefore, using older historical data as samples may not accurately reflect the current user situation, resulting in biased prediction results.

[0004] Secondly, as the market environment changes, the types and structures of product channels may also undergo significant changes. These changes will also affect user acquisition and activity, thereby impacting the TAD prediction results.

[0005] Furthermore, the overall user TAD (Tag and Disability) profile may differ significantly from that of a year ago. This could be due to a combination of factors, including changes in the user base, evolving market trends, and strategic adjustments by competitors. Therefore, using data from a longer period as a sample for prediction may fail to reflect these changes, leading to inaccurate forecasts.

[0006] In view of the above problems, there is an urgent need for a more accurate and efficient TAD prediction method to make full use of recent data that better reflects the current status of the product as a sample for prediction. This application proposes a method and system for predicting total active days of users based on step-by-step supervised learning, aiming to solve the problems existing in traditional methods and improve the accuracy of prediction. Summary of the Invention

[0007] The purpose of this invention is to provide a method and system for predicting total active days of users based on distributed supervised learning. The aim is to maximize the use of recent data that better reflects the current status of the product as a sample for prediction through distributed prediction, so as to effectively improve the accuracy of prediction.

[0008] To achieve the above objectives, the present invention adopts the following technical solution:

[0009] In the first aspect, this application provides a method for predicting the total number of active days of users based on step-by-step supervised learning, including the following steps:

[0010] S1. Group the channel data with new time within a set time window before the target prediction date (e.g., several months before the target prediction date) by month, and extract a multidimensional feature set, including channel attributes, time attributes, TAD values ​​of each stage and TAD ratios of each stage.

[0011] S2. Using the longest period TAD as the target variable and the other features as input features, train the model through a supervised machine learning model to obtain an initial prediction model, and optimize the model parameters based on the prediction error rate index.

[0012] S3. Employing an iterative prediction mechanism, the training samples are gradually expanded to include a dataset containing historical prediction results, and the TAD values ​​for shortening the period are predicted in stages until the TAD values ​​for the entire target prediction period are covered; specific steps include:

[0013] S3.1. During the initial prediction, the longest period TAD is predicted using the initial sample data within the set time window, and the first set of prediction results is generated.

[0014] S3.2 In each subsequent iteration, the prediction results of the previous step are incorporated into the training samples to expand the training dataset;

[0015] S3.3 Based on the expanded dataset, gradually shorten the target TAD cycle and add multi-stage TAD prediction. The target cycle is reduced by a set number of days in each iteration, and at least one new TAD prediction with a shortened cycle is added.

[0016] S3.4 Repeat steps S3.2-S3.3 until the training data covers all shortened phases of the target prediction period (TAD).

[0017] S4. Based on the complete periodic dataset, perform final modeling, extract the TAD value and TAD ratio of all stages, and train to generate the final prediction model.

[0018] S5. Input the newly added channel data into the final prediction model and output the target period TAD prediction result.

[0019] In a preferred embodiment, in step S1, the channel attributes include a unique ID for identifying different channels, the type of the channel, and the number of new channels added.

[0020] The time attribute represents the date of the data record and indicates whether the date is a weekend or a public holiday;

[0021] The TAD values ​​for each stage refer to the total number of active days within different time periods calculated from the date the user is added, including but not limited to 30-day, 60-day, and 90-360-day TADs.

[0022] The TAD ratios for each stage refer to the ratios between TAD values ​​in different stages, including the ratios of TAD values ​​in adjacent periods and combinations of TAD ratios between the initial period and subsequent periods (such as 30 days / 60 days, 30 days / 90 days to 30 days / 360 days, etc.), which are used to reflect the changing trends of user activity.

[0023] In a preferred embodiment, in step S2, the supervised machine learning model is a LightGBM model or an XGBoost model.

[0024] In a preferred embodiment, in step S2, the prediction error rate index is the mean absolute percentage error (MAPE).

[0025] In a preferred embodiment, in step S3.3, the target period is reduced by 30 days each time, and the number of newly added shortened period TAD predictions increases by one in each iteration.

[0026] In a preferred embodiment, in step S4, the full-stage TAD value includes the number of active days over a 360-day period from the date the user is added, and the TAD ratio includes the TAD growth rate over consecutive days.

[0027] In a preferred embodiment, in step S5, the newly added channel data must meet the data accumulation condition of at least 3 days before it can be input into the final prediction model for target period TAD prediction.

[0028] Secondly, this application provides a user total active days prediction system based on step-by-step supervised learning, used to implement the user total active days prediction system based on step-by-step supervised learning as described in any of the first aspects, the system comprising the following modules:

[0029] The data preprocessing module is used to group the channel data within a set time window before the target prediction date by month and extract a multidimensional feature set, which includes channel attributes, time attributes, TAD values ​​of each stage and TAD ratios of each stage.

[0030] The initial modeling module is used to train a supervised machine learning model with the longest period TAD as the target variable and the other features as input features to generate an initial prediction model, and optimize the model parameters based on the prediction error rate metric.

[0031] The iterative prediction module is used to execute an iterative prediction mechanism, progressively expanding the training samples to a dataset containing historical prediction results, predicting the TAD value of the shortened period in stages until the TAD of the entire target prediction period is covered; the iterative prediction module includes:

[0032] The initial prediction unit is used to predict the longest period TAD using initial sample data within a set time window during the first prediction and to generate the first set of prediction results;

[0033] The data expansion unit is used to incorporate the prediction results of the previous step into the training samples in each subsequent iteration, thereby expanding the training dataset.

[0034] The cycle adjustment unit is used to gradually shorten the target TAD cycle based on the expanded dataset and add multi-stage TAD predictions. The target cycle is reduced by a set number of days in each iteration, and at least one new TAD prediction with a shortened cycle is added.

[0035] A cyclic control unit is used to repeatedly call the data expansion unit and the period adjustment unit until the training data covers all shortened phases (TAD) of the target prediction period.

[0036] The final modeling module is used to perform final modeling based on the complete periodic dataset, extract the TAD values ​​and TAD ratios for all stages, and train and generate the final prediction model.

[0037] The prediction output module is used to input new channel data into the final prediction model and output the target period TAD prediction result.

[0038] In a preferred embodiment, the data preprocessing module includes:

[0039] The channel attributes include a unique ID used to identify different channels, the type of the channel, and the number of new channels added.

[0040] The time attribute includes the date of the data record and an identifier indicating whether it is a weekend or a holiday;

[0041] The TAD values ​​for each stage include the total number of active days within different time periods calculated from the date the user is added, including but not limited to 30-day, 60-day, and 90-360-day TADs;

[0042] The TAD ratio for each stage includes the ratio of TAD in adjacent periods and the combination of TAD ratios between the initial period and subsequent periods, used to reflect the changing trend of user activity.

[0043] In a preferred embodiment, the supervised machine learning model in the initial modeling module is a LightGBM model or an XGBoost model.

[0044] In a preferred embodiment, in the initial modeling module, the prediction error rate index is the mean absolute percentage error (MAPE).

[0045] In a preferred embodiment, in the period adjustment unit, the target period is reduced by 30 days each time, and the number of newly added shortened period TAD predictions increases by one in each iteration.

[0046] In a preferred embodiment, in the final modeling module, the full-stage TAD value includes the number of active days over a 360-day period from the date the user is added, and the TAD ratio includes the TAD growth rate over consecutive days.

[0047] In a preferred embodiment, in the prediction output module, the newly added channel data must meet the data accumulation condition of at least 3 days before it can be input into the final prediction model for target period TAD prediction.

[0048] Compared with the prior art, the technical solution of the present invention has the following beneficial effects:

[0049] This invention provides a method and system for predicting total active days (TAD) of users based on step-by-step supervised learning. The method includes: processing channel data within a set time window by month and extracting multi-dimensional feature sets; performing initial training and optimization using a supervised machine learning model; gradually expanding the training samples through an iterative prediction mechanism, predicting the TAD value of shortened cycles in stages until the entire stage is covered; and finally, constructing a prediction model based on the complete dataset to achieve accurate prediction of newly added channel data. This invention maximizes the use of recent data as samples through distributed prediction, more accurately reflecting the current state of the product and user activity patterns, thereby significantly improving prediction accuracy. Furthermore, the step-by-step prediction strategy allows the model to gradually adapt to data changes, maintaining high prediction performance even with frequent product iterations and significant changes in channel structure. Attached Figure Description

[0050] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0051] Figure 1 is a flowchart illustrating the method for predicting total active days of users based on step-by-step supervised learning according to the present invention.

[0052] Figure 2 is a staged prediction data table of user active days (TAD) shown in a preferred embodiment of the present invention. Detailed Implementation

[0053] To make the above and other features and advantages of the present invention clearer, the invention will be further described below with reference to the accompanying drawings. It should be understood that the specific embodiments given herein are for the purpose of explanation to those skilled in the art and are exemplary only, not restrictive.

[0054] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this invention, "a plurality of" means at least two, such as two, three, etc., unless otherwise explicitly specified. Example

[0055] Referring to Figure 1, this embodiment provides a method for predicting the total number of active days of users based on step-by-step supervised learning, which specifically includes the following steps:

[0056] Step S1: Group the channel data with new time within a set time window before the target prediction date (e.g., several months prior to the target prediction date) by month, and extract a multidimensional feature set, including channel attributes, time attributes, TAD values ​​for each stage, and TAD ratios for each stage.

[0057] The channel attributes include a unique ID used to identify different channels, the type of the channel, and the number of new channels added.

[0058] The time attribute represents the date of the data record and indicates whether the date is a weekend or a public holiday;

[0059] The TAD values ​​for each stage refer to the total number of active days within different time periods calculated from the date the user is added, including but not limited to 30-day, 60-day, and 90-360-day TADs.

[0060] The TAD ratios for each stage refer to the ratios between TAD values ​​in different stages, including the ratios of TAD values ​​in adjacent periods and combinations of TAD ratios between the initial period and subsequent periods (such as 30 days / 60 days, 30 days / 90 days to 30 days / 360 days, etc.), which are used to reflect the changing trends of user activity.

[0061] Step S2: Using the longest period TAD (e.g., 360-day TAD) as the target variable and the remaining features as input features, train the model using a supervised machine learning model (e.g., LightGBM or XGBoost) to obtain an initial prediction model, and optimize the model parameters based on the prediction error rate metric (e.g., mean absolute percentage error MAPE).

[0062] Step S3: Employ an iterative prediction mechanism to gradually expand the training samples to include a dataset containing historical prediction results, predicting the TAD value of the shortened period in stages until it covers the TAD of the entire target prediction period; specific steps include:

[0063] Step S3.1: During the initial prediction, the longest period TAD is predicted using the initial sample data within the set time window, and the first set of prediction results is generated.

[0064] Step S3.2: In each subsequent iteration, the prediction results from the previous step are incorporated into the training samples to expand the training dataset;

[0065] Step S3.3: Based on the expanded dataset, gradually shorten the target TAD period (e.g., 330 days, 300 days), and add multi-stage TAD predictions. The target period is reduced by a set number of days in each iteration (e.g., 30 days), and at least one new TAD prediction with a shortened period is added.

[0066] Step S3.4: Repeat steps S3.2-S3.3 until the training data covers all shortened phases of the target prediction period (TAD).

[0067] Step S4: Perform final modeling based on the complete periodic dataset, extract the full-stage TAD value and TAD ratio, and train to generate the final prediction model. The full-stage TAD value includes the total number of active days in the 360-day period from the date the user is added, and the TAD ratio includes the TAD growth rate over consecutive days.

[0068] Step S5: Input the newly added channel data into the final prediction model and output the target period TAD prediction result. The newly added channel data must have accumulated for at least 3 days before being input into the final prediction model for target period TAD prediction.

[0069] To more clearly illustrate the technical solution of this application, a specific example will be used for explanation below.

[0070] Assuming the current date is January 1, 2025, we need to predict the total number of active days (TAD) for each stage of 2024. This embodiment uses a step-by-step supervised learning method to iteratively model using recent data, thereby improving the accuracy of 360-day TAD prediction.

[0071] Step 1: Sample data preparation.

[0072] Time window settings: Select new channel data from October 2023 to the present as the sample, and group them by month (15 months in total). See the data table shown in Figure 2 for details.

[0073] Step 2: Predict the 360-day TAD under the data grouping in January 2024.

[0074] 2.1) Using data from October 2023 to December 2023 as a sample, the following indicators were processed and calculated: Channel ID, Channel Type, New Channel Growth, Whether the Current Date is a Weekend, Whether the Current Date is a Holiday, TAD Values ​​for Each Stage (30 days, 60 days, 90 days...360 days), and TAD Ratios for Each Stage (e.g., 30 days / 60 days, 30 days / 90 days...30 days / 330 days).

[0075] 2.2) After obtaining the above indicators, using the 360-day TAD (longest period) as the target variable and the remaining indicators as feature variables, supervised machine learning models such as LightGBM (lgb) and XGBoost (xgb) are used for training. By comparing the prediction error rate (MAPE) of different models, the model with the smallest error rate and parameters is selected as the optimal model.

[0076] 2.3) Apply the optimal model for prediction: Apply the optimal model and parameters obtained in step 2.2), input the model features under the January 2024 sample, and predict the 360-day TAD for January 2024.

[0077] Step 3: Predict the 330-day TAD and 360-day TAD under the data grouping in February 2024.

[0078] 3.1) Using the prediction results from step two, and taking data from October 2023 to January 2024 as a sample, process and calculate the following indicators: channel ID, channel type, channel new volume, whether the current date is a weekend, whether the current date is a holiday, TAD values ​​for each stage (30 days, 60 days, 90 days...360 days), and TAD ratios for each stage (e.g., 30 days / 60 days, 30 days / 90 days...30 days / 330 days).

[0079] 3.2) After obtaining the above indicators, take 330-day TAD and 360-day TAD as target variables and the remaining indicators as feature variables, call supervised machine learning models such as IGBT and XGBT, and select the model and parameters with the smallest prediction error rate (MAPE).

[0080] 3.3) Apply step 3.2) to obtain the model and parameters, input the model features under the February 2024 sample, and predict and output the 330-day TAD and 360-day TAD for February 2024.

[0081] Step 4: Predict the 300-day TAD, 330-day TAD, and 360-day TAD under the data grouping in March 2024.

[0082] 4.1) Using the prediction results from step three, and taking data from October 2023 to February 2024 as a sample, process and calculate the following indicators: Channel ID, Channel type, Channel new volume, Whether the current date is a weekend, Whether the current date is a holiday, TAD values ​​for each stage (30 days, 60 days, 90 days...360 days), and TAD ratios for each stage (e.g., 30 days / 60 days, 30 days / 90 days...30 days / 270 days).

[0083] 4.2) After obtaining the above indicators, take 300-day TAD, 330-day TAD, and 360-day TAD as target variables and the remaining indicators as feature variables. Call supervised machine learning models such as IGBT and XGBT, and select the model and parameters with the smallest prediction error rate (MAPE).

[0084] 4.3) Apply step 4.2) to obtain the model and parameters, input the model features under the March 2024 sample, and predict and output the 300-day TAD, 330-day TAD, and 360-day TAD for March 2024.

[0085] Step 5: Continue in this manner, repeatedly modeling and predicting the TAD for each stage under the full-year data of 2024 (see the yellow part of the data table in Figure 2).

[0086] By using rolling forecasts, the TAD forecast values ​​for each month of 2024 at different stages (such as 30 days, 60 days... 360 days) will be gradually supplemented.

[0087] Step Six: After completing the steps, perform the final modeling.

[0088] 6.1) Using data from October 2023 to December 2024 as a sample (including the predicted TAD values), the following more comprehensive indicators were processed and calculated: Channel ID, Channel type, Channel new volume, whether the current date is a weekend, whether the current date is a holiday, TAD values ​​for each stage (1 day, 2 days, 3 days...360 days) and TAD ratios for each stage (such as 2 days / 1 day, 3 days / 1 day, etc.).

[0089] 6.2) After obtaining the above indicators, take 360-day TAD as the target variable and the other indicators as feature variables, call supervised machine learning models such as IGBT and XGBT, select the model and parameters with the smallest prediction error rate (MAPE), and train to generate the final prediction model.

[0090] Step 7: Predicting new channels.

[0091] Using the final prediction model and parameters obtained in step six, and assuming that the newly added channels have been active for 3 days, input the model features (including channel ID, channel type, number of new channels, whether the current date is a weekend, whether the current date is a holiday, and the calculated TAD value, etc.) to predict the 360-day TAD of the channel (see the blue part of the data table in Figure 2).

[0092] In summary, this application employs an iterative mechanism that effectively integrates recent data with prediction results. Each iteration incorporates the latest prediction results into the training samples, resolving the issue of historical data lag and ensuring that the model is always optimized based on the current product status (e.g., prediction results from January 2024 are used in model building for February). This application further shortens the target prediction period (reducing it by 30 days each time) and adds prediction phases (e.g., the first prediction is 360 days, the second is 330 days, and the third is 360 days), enabling the model to simultaneously capture short-term and long-term user activity trends.

[0093] Compared to traditional methods, the technical solution in this embodiment significantly reduces reliance on outdated data. Traditional methods typically rely on historical data from 360 days ago, while the technical solution in this embodiment only requires three months of initial data (October to December 2023) to start prediction. Subsequent iterations expand the sample, reducing reliance on outdated data. In this embodiment, the prediction for January 2024 uses only three months of data, gradually expanding to full-year data, resulting in more reasonable resource utilization. For new channels, only three days of data need to be accumulated to predict 360 days of user activity days (TAD) through the final constructed model, thus meeting the real-time needs of the business. Example

[0094] Based on the same design concept, this embodiment also provides a user total active days prediction system based on step-by-step supervised learning, used to execute the user total active days prediction method based on step-by-step supervised learning described in Embodiment 1.

[0095] Specifically, the system includes: a data preprocessing module, an initial modeling module, an iterative prediction module, a final modeling module, and a prediction output module.

[0096] The data preprocessing module is used to group the channel data within a set time window before the target prediction date by month and extract a multidimensional feature set, which includes channel attributes, time attributes, TAD values ​​for each stage, and TAD ratios for each stage.

[0097] The channel attributes include a unique ID for identifying different channels, the type of channel, and the number of new users added to the channel; the time attributes include the date of the data record and whether it is a weekend or a holiday; the TAD values ​​for each stage include the total number of active days in different time periods calculated from the date of user addition, including but not limited to 30-day, 60-day, 90-day to 360-day TADs; the TAD ratios for each stage include the ratio of TADs in adjacent periods and the combination of the TAD ratios of the initial period and subsequent periods, used to reflect the changing trend of user activity.

[0098] The initial modeling module is used to train a supervised machine learning model with the longest period TAD as the target variable and the remaining features as input features to generate an initial prediction model, and optimize the model parameters based on the prediction error rate metric. In a preferred embodiment, the supervised machine learning model is a LightGBM model or an XGBoost model; the prediction error rate metric is the mean absolute percentage error (MAPE).

[0099] The iterative prediction module is used to execute an iterative prediction mechanism, gradually expanding the training samples to a dataset containing historical prediction results, predicting the TAD value of the shortened period in stages, until the TAD of the entire target prediction period is covered.

[0100] Specifically, the iterative prediction module includes:

[0101] The initial prediction unit is used to predict the longest period TAD using initial sample data within a set time window during the first prediction and to generate the first set of prediction results;

[0102] The data expansion unit is used to incorporate the prediction results of the previous step into the training samples in each subsequent iteration, thereby expanding the training dataset.

[0103] The cycle adjustment unit is used to gradually shorten the target TAD cycle based on the expanded dataset and add multi-stage TAD predictions. The target cycle is reduced by a set number of days (e.g., 30 days) in each iteration, and at least one new TAD prediction with a shortened cycle is added.

[0104] The loop control unit is used to repeatedly call the data expansion unit and the period adjustment unit until the training data covers all shortened phases (TAD) of the target prediction period.

[0105] The final modeling module is used to perform final modeling based on the complete periodic dataset, extracting the full-stage TAD value and TAD ratio, and training to generate the final prediction model. The full-stage TAD value includes the total number of active days over the 360-day period from the date the user is added, and the TAD ratio includes the TAD growth rate over consecutive days.

[0106] The prediction output module is used to input new channel data into the final prediction model and output the target period TAD prediction result. The new channel data must have accumulated for at least 3 days before it can be input into the final prediction model for target period TAD prediction.

[0107] It is understood that the modules described in this step-by-step supervised learning-based total active days prediction system correspond to the steps in the step-by-step supervised learning-based total active days prediction method described in Figure 1. Therefore, the operations, features, and beneficial effects described above for the step-by-step supervised learning-based total active days prediction method also apply to the step-by-step supervised learning-based total active days prediction system and its included modules, and will not be repeated here.

[0108] The specific embodiments of the present invention have been described in detail above, but they are merely examples, and the present invention is not limited to the specific embodiments described above. For those skilled in the art, any equivalent modifications and substitutions to the present invention are also within the scope of the present invention. Therefore, all equivalent transformations and modifications made without departing from the spirit and scope of the present invention should be covered within the scope of the present invention.

Claims

1. A method for predicting total user activity days based on step-by-step supervised learning, characterized in that, Includes the following steps: S1. Group the channel data within the time window set before the target prediction date by month and extract a multidimensional feature set, including channel attributes, time attributes, TAD values ​​of each stage and TAD ratios of each stage. S2. Using the longest period TAD as the target variable and the other features as input features, train the model through a supervised machine learning model to obtain an initial prediction model, and optimize the model parameters based on the prediction error rate index. S3. An iterative prediction mechanism is adopted to gradually expand the training samples to a dataset containing historical prediction results, and predict the TAD value of the shortened period in stages until the TAD of the target prediction period is covered. The specific steps include: S3.

1. During the initial prediction, the longest period TAD is predicted using the initial sample data within the set time window, and the first set of prediction results is generated. S3.2 In each subsequent iteration, the prediction results of the previous step are incorporated into the training samples to expand the training dataset; S3.3 Based on the expanded dataset, gradually shorten the target TAD cycle and add multi-stage TAD prediction. The target cycle is reduced by a set number of days in each iteration, and at least one new TAD prediction with a shortened cycle is added. S3.4 Repeat steps S3.2-S3.3 until the training data covers all shortened phases of the target prediction period (TAD). S4. Based on the complete periodic dataset, perform final modeling, extract the TAD value and TAD ratio of all stages, and train to generate the final prediction model. S5. Input the newly added channel data into the final prediction model and output the target period TAD prediction result.

2. The method for predicting total user activity days based on step-by-step supervised learning according to claim 1, characterized in that, In step S1, the channel attributes include a unique ID used to identify different channels, the type of the channel, and the number of new channels added. The time attribute represents the date of the data record and indicates whether the date is a weekend or a public holiday; The TAD value for each stage refers to the total number of active days within different time periods, calculated from the date the user is added. The TAD ratio for each stage refers to the ratio between TAD values ​​in different stages, including the ratio of TAD values ​​in adjacent periods and the combination of TAD ratios in the initial period and subsequent periods, which is used to reflect the changing trend of user activity.

3. The method for predicting total user activity days based on step-by-step supervised learning according to claim 1, characterized in that, In step S2, the supervised machine learning model is either the LightGBM model or the XGBoost model.

4. The method for predicting total user activity days based on step-by-step supervised learning according to claim 1, characterized in that, In step S2, the prediction error rate index is the mean absolute percentage error (MAPE).

5. The method for predicting total user activity days based on step-by-step supervised learning according to claim 1, characterized in that, In step S3.3, the target period is reduced by 30 days each time, and the number of newly added shortened period TAD predictions increases by one in each iteration.

6. The method for predicting total user activity days based on step-by-step supervised learning according to claim 1, characterized in that, In step S4, the full-stage TAD value includes the number of active days over a 360-day period from the date the user is added, and the TAD ratio includes the TAD growth rate over consecutive days.

7. The method for predicting total user activity days based on step-by-step supervised learning according to claim 1, characterized in that, In step S5, the newly added channel data must meet the data accumulation condition of at least 3 days before it can be input into the final prediction model for target period TAD prediction.

8. A user total active days prediction system based on step-by-step supervised learning, characterized in that, Includes the following modules: The data preprocessing module is used to group the channel data within a set time window before the target prediction date by month and extract a multidimensional feature set, which includes channel attributes, time attributes, TAD values ​​of each stage and TAD ratios of each stage. The initial modeling module is used to train a supervised machine learning model with the longest period TAD as the target variable and the other features as input features to generate an initial prediction model, and optimize the model parameters based on the prediction error rate metric. The iterative prediction module is used to execute the iterative prediction mechanism, gradually expanding the training samples to a dataset containing historical prediction results, predicting the TAD value of the shortened period in stages, until the TAD of the entire target prediction period is covered; The iterative prediction module includes: The initial prediction unit is used to predict the longest period TAD using initial sample data within a set time window during the first prediction and to generate the first set of prediction results; The data expansion unit is used to incorporate the prediction results of the previous step into the training samples in each subsequent iteration, thereby expanding the training dataset. The cycle adjustment unit is used to gradually shorten the target TAD cycle based on the expanded dataset and add multi-stage TAD predictions. The target cycle is reduced by a set number of days in each iteration, and at least one new TAD prediction with a shortened cycle is added. A cyclic control unit is used to repeatedly call the data expansion unit and the period adjustment unit until the training data covers all shortened phases (TAD) of the target prediction period. The final modeling module is used to perform final modeling based on the complete periodic dataset, extract the TAD values ​​and TAD ratios for all stages, and train and generate the final prediction model. The prediction output module is used to input new channel data into the final prediction model and output the target period TAD prediction result.

9. The user total active days prediction system based on step-by-step supervised learning according to claim 8, characterized in that, In the data preprocessing module: The channel attributes include a unique ID used to identify different channels, the type of the channel, and the number of new channels added. The time attribute includes the date of the data record and an identifier indicating whether it is a weekend or a holiday; The TAD values ​​for each stage include the total number of active days within different time periods, calculated from the date the user is added. The TAD ratio for each stage includes the ratio of TAD in adjacent periods and the combination of TAD ratios between the initial period and subsequent periods, used to reflect the changing trend of user activity.

10. The user total active days prediction system based on step-by-step supervised learning according to claim 8, characterized in that, In the final modeling module, the full-stage TAD value includes the number of active days over a 360-day period from the date the user is added, and the TAD ratio includes the TAD growth rate over consecutive days.

Images