A method for managing book order data

Abstract

The present invention provides a method for managing book order data, which relates to the technical field of data processing. The method includes: collecting historical book order data of a platform, constructing a plurality of book order feature matrices, and training an order time feature analysis model for each type of book; collecting platform user data corresponding to the historical book order data, constructing a platform user feature matrix, and training a platform user feature analysis model; generating order prediction parameters of multiple time features through the order time feature analysis model corresponding to each type of book; generating user prediction parameters of multiple time features through the platform user feature analysis model; extracting order-user association parameters of each type of book under multiple time features, calculating priority parameters of each type of book under multiple time features, and adjusting the recommendation strategy of multiple types of books on the platform based on the priority parameters. The present invention provides a basis for optimizing the book recommendation strategy of the platform.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and particularly relates to a method for managing book order data. Background Art

[0002] With the rapid development of e-commerce platforms, the book market has gradually shifted towards online sales, and consumers' online book purchases have become a common way of buying books. Due to the wide variety of books on online platforms and diverse user needs, how to effectively manage and analyze book order data has become the key to enhancing the competitiveness of the platform. Through in-depth analysis of book order data, the platform can not only understand the sales trends of different books but also uncover potential user demand information. This information is crucial for formulating precise marketing strategies, optimizing inventory management, and enhancing the user experience.

[0003] The traditional way of adjusting the book recommendation strategy of the platform based on a single dimension such as the category and sales volume of books can, to a certain extent, meet the basic needs of users. However, due to the lack of in-depth analysis of user behavior on the platform and the failure to effectively combine key factors such as user activity and order conversion rate, there is a deviation between the recommended content and the actual needs of users, affecting the accuracy of recommendations and the user experience. In addition, most of the existing book order data management methods focus on the processing of static data, ignoring the impact of the time dimension on user purchase behavior and book sales, and the level of dynamic analysis and prediction of book orders and user behavior needs to be improved. Summary of the Invention

[0004] To solve the above technical problems, the present invention proposes a method for managing book order data, aiming to deeply mine the book order data and user data of the platform, identify the personalized needs of platform users, provide a basis for optimizing the book recommendation strategy of the platform, and facilitate the platform to achieve personalized book recommendations.

[0005] To achieve the above object, the present invention provides a method for managing book order data, including:

[0006] Collect historical book order data of the platform, extract the order quantity characteristics and order time characteristics of each type of book from the historical book order data, and construct multiple book order feature matrices based on the order quantity characteristics and order time characteristics of each type of book;

[0007] Input the book order feature matrix into the order time feature analysis model, use the order time feature as the input feature of the order time feature analysis model, and use the order quantity feature as the label to train the order time feature analysis model for each type of book and generate the target order time feature for each type of book;

[0008] Collect the platform user data corresponding to the historical book order data, extract the user quantity feature and user time feature from the platform user data, construct a platform user feature matrix based on the user quantity feature and user time feature, input the platform user feature matrix into the platform user feature analysis model, use the user time feature as the input feature of the platform user feature analysis model, use the user quantity feature as the label, train the platform user feature analysis model and generate the target user time feature of the platform;

[0009] Input the target user time feature and the target order time feature of each type of book into the order time feature analysis model corresponding to each type of book to generate order prediction parameters for multiple time features;

[0010] Input the target user time feature and the target order time feature of each type of book into the platform user feature analysis model to generate user prediction parameters for multiple time features;

[0011] Extract the order-user association parameters of each type of book under the target order time feature and target user time feature from the historical book order data and platform user data, calculate the priority parameters of each type of book under multiple time features according to the order prediction parameters, user prediction parameters and order-user association parameters, and adjust the recommendation strategy of multiple types of books on the platform based on the priority parameters.

[0012] Preferably, extract the order-user association parameters of each type of book under the target order time feature and target user time feature from the historical book order data and platform user data, including:

[0013] Extract the order quantity data of each type of book under each target order time feature and target user time feature from the historical book order data, and extract the platform active user data under each target order time feature and target user time feature from the platform user data;

[0014] Calculate the order-user association parameter of each type of book under each target order time feature and target user time feature according to the order quantity data and platform active user data, and the order-user association parameter is the ratio of the order quantity to the platform active users.

[0015] Preferably, calculate the priority parameters of each type of book under multiple time features according to the order prediction parameters, user prediction parameters and order-user association parameters, including:

[0016] Process the order prediction parameters and user prediction parameters under each time feature based on a preset time feature unit to obtain order normalization parameters and user normalization parameters, and determine the recommendation identifier under each time feature according to the order normalization parameters;

[0017] Associate the order standardization parameters, user standardization parameters, order-user association parameters, and recommendation identifiers under multiple time characteristics and construct an order feature dataset;

[0018] Input the order feature dataset into the associated feature analysis model. Use the order standardization parameters, user standardization parameters, and order-user association parameters in the order feature dataset as the input features of the associated feature analysis model, and use the recommendation identifier as the label. Through the associated feature analysis model, respectively output the target scores corresponding to the order standardization parameters, user standardization parameters, and order-user association parameters, and calculate the priority parameters of each type of book under multiple time characteristics based on the target scores;

[0019] Among them, calculating the priority parameters of each type of book under multiple time characteristics based on the target scores includes determining the weight parameters corresponding to the order standardization parameters, user standardization parameters, and order-user association parameters according to the target scores, and performing weighted calculations on the order standardization parameters, user standardization parameters, and order-user association parameters under each time characteristic based on the weight parameters to obtain the priority parameters of each type of book under multiple time characteristics.

[0020] Preferably, based on a preset time feature unit, process the order prediction parameters and user prediction parameters under each time characteristic to obtain the order standardization parameters and user standardization parameters, and determine the recommendation identifier under each time characteristic according to the order standardization parameters, including:

[0021] Based on the preset time feature unit, perform normalization processing on each time characteristic, calculate the normalization parameter of each time characteristic, and correct the order prediction parameters and user prediction parameters under each time characteristic according to the normalization parameter to obtain the order standardization parameters and user standardization parameters under each time characteristic. Determine the recommendation identifier under each time characteristic based on a preset quantity threshold and the order standardization parameters, including recording the recommendation identifier under the time characteristic where the order standardization parameter is greater than the preset quantity threshold as 1, otherwise as 0.

[0022] Preferably, for the book order feature matrix and the platform user feature matrix, it further includes:

[0023] Extract the time series data of each type of book order from the historical book order data and construct an order time series dataset. Train an order time series analysis model through the time series dataset. Predict the order time series prediction quantity of each type of book in the first preset period through the order time series analysis model. Determine the order time characteristics corresponding to the first preset period and the order quantity characteristics corresponding to the order time series prediction quantity in the first preset period. Add the order time characteristics corresponding to the first preset period and the order quantity characteristics corresponding to the order time series prediction quantity in the first preset period to the book order feature matrix corresponding to the book order of the corresponding type. Train an order time feature analysis model for each type of book through the book order feature matrix;

[0024] Extract the time series data of the user quantity from the platform user data and construct a user time series dataset. Train a user time series analysis model through the user time series dataset. Predict the user time series prediction quantity in the second preset period through the order time series analysis model. Determine the user time characteristics corresponding to the second preset period and the user quantity characteristics corresponding to the user time series prediction quantity in the second preset period. Add the user time characteristics corresponding to the second preset period and the user quantity characteristics corresponding to the user time series prediction quantity in the second preset period to the platform user feature matrix. Train a platform user feature analysis model based on the platform user feature matrix.

[0025] Preferably, select the order time characteristics of each type of book with multiple feature importance scores greater than the preset scoring threshold as the target order time characteristics of each type of book, and select the user time characteristics with multiple feature importance scores greater than the preset scoring threshold as the target user time characteristics of the platform.

[0026] Preferably, the feature analysis model is constructed based on the random forest model, and the time series analysis model is constructed based on the ARIMA model.

[0027] The present invention has the following beneficial effects:

[0028] By analyzing data across multiple time dimensions, the present invention can more comprehensively capture the characteristics of order quantity and user activity changes over time, thereby providing more accurate predictions and recommendations within different time periods. By fully leveraging the time characteristics implicit in historical data and introducing future data predicted by time series models, it can not only capture the static time characteristics in historical data but also identify trend and periodic changes in the time series. The data output by the time series model provides predictions of future order volume and user activity, enabling the random forest model to make more accurate decisions based on this forward-looking information, helping the model better understand the complex dependencies of time characteristics. The introduction of dynamic information enables the model to better respond to complex market changes and improves the overall prediction accuracy. The platform can adjust the book recommendation strategy in real time based on the calculated priority parameters, facilitating the formulation of a more refined and personalized book recommendation strategy to meet the needs of different time periods and user groups. Relevant books are preferentially recommended during peak periods of order volume or user activity, enhancing the relevance and user satisfaction of the recommendations, and thereby improving the user experience and conversion rate. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 FIG. is a schematic flowchart of a method for managing book order data provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0030] In order to enable those skilled in the art to better understand the technical solutions in the present invention, the following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0031] As Figure 1 shown, a method for managing book order data provided in one embodiment of the present invention includes the following steps:

[0032] S1. Collect historical book order data of the platform, extract the order quantity characteristics and order time characteristics of each type of book from the historical book order data, and construct multiple book order feature matrices based on the order quantity characteristics and order time characteristics of each type of book.

[0033] Specifically, data related to book orders is extracted from the historical data of the platform. The collected historical book order data usually includes the sales records of each type of book in different time periods. The data collected from the platform includes, but is not limited to, information such as the book categories involved in the book orders, order timestamps, and order quantities, to characterize the historical sales situation of books on the platform. The order quantity feature and order time feature of each type of book are extracted from the collected data. The order quantity feature can include data such as the total order quantity and average order quantity of this type of book in different time periods. The order time feature can include the specific time features when the orders occur, such as the order distribution of a certain type of book in a day and the order distribution situation in a week. The order time feature can specifically be data in multiple dimensions such as hours, days, weeks, and months. The book order feature matrix constructed based on the order quantity feature and order time feature, where the rows can represent the order time features in multiple dimensions, and the columns can represent the corresponding order quantity features in each time dimension.

[0034] S2. Input the book order feature matrix into the order time feature analysis model. Using the order time feature as the input feature of the order time feature analysis model and the order quantity feature as the label, train the order time feature analysis model for each type of book and generate the target order time feature for each type of book.

[0035] Specifically, the goal of the order time feature analysis model is to capture and learn the relationship between the time feature and the order quantity. After receiving the input book order feature matrix, the model uses the order time feature in the book order feature matrix as the input feature and the order quantity feature in the book order feature matrix as the label, that is, the target value that the model needs to predict, such as the order quantity of a certain type of book under the feature order time feature. By training the order time feature analysis model with the book order feature matrix, finally, the order time feature analysis model for each type of book can be obtained. The order time feature analysis model for each type of book can output the target order time feature of the corresponding type of book, that is, predict the possible order quantity of this type of book in different future time periods.

[0036] In this embodiment, the feature analysis model is specifically a random forest model. Random forest is an ensemble learning method that consists of many decision trees. By combining the prediction results of multiple decision trees, it can improve the accuracy and stability of the model and automatically handle the correlation between input features and labels. Compared with other models, after training, the random forest model can calculate the importance scores of each input feature (such as hour, day of the week, month). These scores represent the degree of influence of each time feature on the number of book orders. Thus, according to the importance scores output by the model, the order time features with a greater degree of influence on the order quantity can be selected from multiple order time features, denoted as target order time features. Through the random forest model, it can be identified which time periods have a significant impact on the sales of each type of book. For example, it may be found that the sales peak of children's books is after school in the afternoon, while the sales peak of academic books is in the evenings on weekdays.

[0037] It should be noted that during the process of training the order time feature analysis model through the book order feature matrix, the model will mine the correlation between input features and labels in the data. Since the historical data of different types of books are different, in order to improve the training accuracy, an order time feature analysis model is trained separately for each type of book instead of inputting all the data into a single model for training. In this case, the model can more accurately fit the relationship between the order quantity of this type of book and time features, avoiding noise interference between different types of books.

[0038] S3. Collect the platform user data corresponding to the historical book order data, extract the user quantity feature and user time feature from the platform user data, and construct a platform user feature matrix based on the user quantity feature and user time feature;

[0039] Specifically, from the historical book order data of the platform, the corresponding platform user data is associated. The platform user data is specifically the active user data of the platform during the time corresponding to the historical book order data. For example, if the historical book order data is the book order - related data of the platform at different time periods within the past three months, then the platform user data is the active user - related data of the platform at different time periods within the past three months. The user quantity feature and user time feature are extracted from the platform user data. Among them, the user quantity feature corresponds to the user time feature. For example, if the user time feature is a certain hour of the day or a certain day of the week, there is a corresponding user quantity feature, such as the number of active users on the content platform at a certain hour of the day or the total number of active users on the platform within a certain day of the week. A platform user feature matrix is constructed based on the extracted user quantity feature and user time feature. Among them, the rows of the matrix can represent different user time features, and the columns of the matrix correspondingly represent the number of active users on the platform under each user time feature.

[0040] It should be noted that the user time feature and the order time feature can be determined using the same time division method to extract rich feature information in different dimensions. In this embodiment, the division methods of the user time feature and the order time feature are not specifically limited.

[0041] S4. Input the platform user feature matrix into the platform user feature analysis model. Use the user time feature as the input feature of the platform user feature analysis model and the user quantity feature as the label to train the platform user feature analysis model and generate the target user time feature of the platform.

[0042] Specifically, input the constructed user feature matrix into the platform user feature analysis model for model training. The goal of the platform user feature analysis model is to capture the time regularity of user behavior and predict the user activity level at different time periods based on this. Use the user time feature as the input feature and the user quantity feature as the target value to be predicted by the model. Finally, the model can be used to predict the number of active users on the platform at different time periods. Similarly, after the training of the platform user feature analysis model, it also outputs the importance scores of different user time features, so that multiple target user time features with a greater impact on the number of active users on the platform can be determined according to the importance scores of different user time features.

[0043] In this embodiment, for the target order time feature and the target user time feature, they can be filtered based on a preset scoring threshold. For example, filter out the order time features of each type of book whose multiple feature importance scores are greater than the preset scoring threshold to obtain the target order time feature of each type of book, and filter out the user time features whose multiple feature importance scores are greater than the preset scoring threshold to obtain the target user time feature of the platform.

[0044] S5. Input the target user time feature and the target order time feature of each type of book into the order time feature analysis model corresponding to each type of book to generate order prediction parameters for multiple time features. Input the target user time feature and the target order time feature of each type of book into the platform user feature analysis model to generate user prediction parameters for multiple time features.

[0045] Specifically, the target user time features obtained in step S4 and the target order time features of each type of book generated in step S2 are respectively input into the order time feature analysis model of each type of book to generate order prediction parameters under multiple time features, specifically, order prediction parameters corresponding to each type of book under multiple time features. The order prediction parameter represents the predicted order quantity of each type of book during these time periods. At the same time, the target user time features and the target order time features of each type of book are input into the platform user feature analysis model to generate user prediction parameters under multiple time features. The user prediction parameter represents the predicted number of active users on the platform during these time periods.

[0046] It should be noted that the target user time features and the target order time features respectively characterize multiple time periods that have a greater impact on the book order quantity and the number of active users on the platform. Through the above method, the order quantity and the number of active users on the platform can be associated, and the relationship between the number of active users and the order quantity can be analyzed from the order perspective, as well as the relationship between the order quantity and the active users can be analyzed from the active users.

[0047] S6. Extract the order-user association parameters of each type of book under the target order time features and the target user time features from the historical book order data and the platform user data. Calculate the priority parameters of each type of book under multiple time features according to the order prediction parameters, the user prediction parameters, and the order-user association parameters, and adjust the recommendation strategy of multiple types of books on the platform based on the priority parameters.

[0048] Specifically, from the historical book order data and the platform user data, extract the order-user association parameters of each type of book under the target order time features and the target user time features. The order-user association parameter is used to quantify the relationship between user activity and order quantity at different time periods.

[0049] In one implementation process, extracting the order-user association parameters of each type of book under the target order time features and the target user time features from the historical book order data and the platform user data specifically includes:

[0050] Extract the order quantity data of each type of book under each target order time feature and the target user time feature from the historical book order data, and extract the platform active user data under each target order time feature and the target user time feature from the platform user data.

[0051] Specifically, the order quantity data reflects the actual sales of different types of books under specific time characteristics (such as a certain time period of a day, a certain day of the week, or a specific holiday), which can intuitively reflect the market performance and sales trends of each type of book under different time characteristics. The active user data reflects how many users on the platform are active under the same time characteristics, specifically the number of active users. Among them, active users can refer to users who visit the platform within a certain time period, users who browse a certain type of book, users who add items to the shopping cart or place an order, etc.

[0052] After obtaining the order quantity data and the platform active user data, calculate the order-user association parameter of each type of book under each target order time characteristic and target user time characteristic according to the order quantity data and the platform active user data. The order-user association parameter is used to quantify the relationship between user activity and order quantity. In this embodiment, the order-user association parameter is represented by the ratio of the order quantity to the number of active users on the platform. By analyzing the association parameters under different time characteristics, it can be identified which time periods have a higher user conversion rate.

[0053] After obtaining the order-user association parameter of each type of book, combine the order prediction parameter, user prediction parameter, and order-user association parameter to calculate the priority parameter of each type of book under multiple time characteristics. The priority parameter reflects the recommended priority of each type of book in different time periods. Based on the calculated priority parameter, the recommendation strategy of multiple types of books on the platform can be dynamically adjusted. Exemplarily, those books with higher priority parameters are preferentially recommended in different time periods to improve the relevance and user satisfaction of the recommendation.

[0054] In one implementation process, in step S6, calculating the priority parameter of each type of book under multiple time characteristics according to the order prediction parameter, user prediction parameter, and order-user association parameter specifically includes:

[0055] S61. Process the order prediction parameter and user prediction parameter under each time characteristic based on a preset time characteristic unit to obtain an order normalization parameter and a user normalization parameter, and determine the recommendation identifier under each time characteristic according to the order normalization parameter.

[0056] Specifically, considering the different magnitude differences of the order prediction parameter and user prediction parameter under different time scales, the order prediction parameter and user prediction parameter are normalized through a preset time characteristic unit. Specifically, each time characteristic is normalized based on the preset time characteristic unit, and the normalization parameter of each time characteristic is calculated. The order prediction parameter and user prediction parameter under each time characteristic are corrected according to the normalization parameter to obtain the order normalization parameter and user normalization parameter under each time characteristic.

[0057] Exemplarily, taking 1 hour as the preset time feature unit, for the time feature of one day, the normalization parameter is 24. The order prediction parameter and user prediction parameter under each time feature are corrected according to the normalization parameter. For example, calculate the ratio of the total number of active users and the total number of orders on the platform within one day to the normalization parameter, so as to obtain the number of active users and the number of orders within the scale of the preset time feature unit, that is, the order normalization parameter and user normalization parameter under each time feature. And based on the preset quantity threshold and the order normalization parameter, the recommendation flag for each time feature is determined. The recommendation flag is used to indicate whether a certain type of book has a high recommendation priority under a specific time feature. In this embodiment, the recommendation flag under the time feature where the order normalization parameter is greater than the preset quantity threshold is recorded as 1, otherwise it is recorded as 0.

[0058] S62. Associate the order normalization parameters, user normalization parameters, order-user association parameters, and recommendation flags under multiple time features and construct an order feature data set. The order feature data set combines the information of orders, users, and order-user associations, providing complete feature inputs for the training of the model.

[0059] S63. Input the order feature data set into the associated feature analysis model. Use the order normalization parameter, user normalization parameter, and order-user association parameter in the order feature data set as the input features of the associated feature analysis model, and use the recommendation flag as the label. Through the associated feature analysis model, the target scores corresponding to the order normalization parameter, user normalization parameter, and order-user association parameter are respectively output, and the priority parameter of each type of book under multiple time features is calculated based on the target scores.

[0060] Specifically, input the order feature data set into the associated feature analysis model for training. The goal of the model is to learn the relationship between the order normalization parameter, user normalization parameter, and order-user association parameter and the recommendation flag, and identify the influence degree of these three on the recommendation flag. Specifically, the model generates the target scores corresponding to the order normalization parameter, user normalization parameter, and order-user association parameter respectively, which are used to quantitatively represent the influence degree of each feature on the recommendation flag.

[0061] After obtaining the target scores corresponding to the order standardization parameters, user standardization parameters, and order-user association parameters respectively, calculate the priority parameters of each type of book under multiple time features based on the target scores. Specifically, use the target scores corresponding to the order standardization parameters, user standardization parameters, and order-user association parameters as their own weight parameters, and perform weighted calculations on the order standardization parameters, user standardization parameters, and order-user association parameters under each time feature based on the weight parameters to obtain the priority parameters of each type of book under multiple time features.

[0062] It should be noted that using the order standardization parameters, user standardization parameters, and order-user association parameters for weighted calculation can comprehensively consider multiple key factors, which is more comprehensive than relying only on a single feature for recommendation and can more accurately reflect the recommendation priority of books within a specific time period. The order standardization parameter reflects the market demand and sales performance of a certain type of book within a specific time period. The user standardization parameter takes into account the fluctuations in user activity on the platform, which affects the exposure opportunities and potential purchase behaviors of books. The order-user association parameter indicates the efficiency of converting active users into actual purchasing users (i.e., the conversion rate). This parameter helps to identify those time periods with high user activity but low conversion rates, so as to improve the recommendation effect targeted. By analyzing the order feature dataset through a random forest model, it is possible to well identify the influence of the order prediction parameter, user prediction parameter, and order-user association parameter in the recommendation under different scenarios, so as to obtain the time period suitable for recommending each type of book, making the recommendation decision for each time period more in line with the actual user behavior and order situation, thereby reducing irrelevant or inefficient recommendations. By calculating the priority parameters in this way, it can be updated regularly based on the collected historical data, which is convenient for the platform to formulate a recommendation strategy suitable for the market according to the actual situation, so that the recommendation strategy can dynamically respond to user behavior and market changes.

[0063] In one implementation process, for the order time feature analysis model and the platform user feature analysis model, it further includes:

[0064] Extract the time series data of each type of book order from the historical book order data and construct an order time series dataset. Train an order time series analysis model through the time series dataset, and predict the order time series prediction quantity of each type of book in the first preset time period through the order time series analysis model. Among them, the time series data of the book order contains the order quantity of this type of book at each time point, captures the periodic, trend, and seasonal change patterns of the order quantity, and reflects the change trend of the order quantity over time. The order time series analysis model is used to learn the law of the order quantity changing over time and is subsequently used to predict the order time series prediction quantity of books in different time periods.

[0065] After obtaining the order time series prediction quantity for the first preset period, determine the order time characteristics corresponding to the first preset period and the order quantity characteristics corresponding to the order time series prediction quantity for the first preset period, and add the order time characteristics corresponding to the first preset period and the order quantity characteristics corresponding to the order time series prediction quantity for the first preset period to the book order feature matrix corresponding to the book order of the corresponding type, so as to update the training data of the order time characteristic analysis model, and train the order time characteristic analysis model for each type of book through the updated book order feature matrix.

[0066] Extract the time series data of the user quantity from the platform user data and construct a user time series data set, train a user time series analysis model through the user time series data set, and predict the user time series prediction quantity for the second preset period through the order time series analysis model. Among them, the time series data of the user quantity records the activity of platform users in different time periods, reflects the time law of platform user behavior, and the order time series analysis model aims to learn the time law of user activity, capture the behavior patterns of users in different time periods, and through analyzing the active time series data of users, the model can identify and predict the periodic changes of user activity, such as the differences between weekdays and weekends, and the increase in activity during holidays.

[0067] After determining the user time series prediction quantity for the second preset period, determine the user time characteristics corresponding to the second preset period and the user quantity characteristics corresponding to the user time series prediction quantity for the second preset period, and add the user time characteristics corresponding to the second preset period and the user quantity characteristics corresponding to the user time series prediction quantity for the second preset period to the platform user feature matrix. By introducing prediction data, the platform user feature matrix not only reflects historical user behavior, but also includes the expected future user activity. Train a platform user feature analysis model based on the platform user feature matrix. In this embodiment, the first preset period and the second preset period can be a future day, week or month, and those skilled in the art can make reasonable settings according to actual needs. For the order time series analysis model and the user time series analysis model, in this embodiment, they are constructed based on the ARIMA model.

[0068] It should be noted that although the book order feature matrix pre-constructed based on historical book order data and the platform user feature matrix constructed based on the corresponding platform user data of historical book order data already contain information on time features in the data, this information is scattered at different time points, and the random forest model mainly looks for static patterns between these time points. For the time series analysis model, although the predicted data for a certain period in the future output can to some extent be regarded as historical data, the predicted output of the time series analysis model is not just a simple repetition of historical data, but a kind of "foresight" of the future. The time series analysis model extracts the trends and changes (such as upward and downward trends) between time points to enhance the input of the random forest model. The data output by the time series model is essentially the order volume or activity at a certain future time point predicted based on historical trends. This provides a forward-looking perspective that allows the random forest model to use this perspective to adjust the current prediction. When the random forest model is trained, if it only relies on the time features in historical data, what it captures may be more of the "point-like relationships" between time points rather than the "continuity" in time. By adding the output of the time series model, the random forest model can learn which combinations of time features are most likely to lead to high order volumes or high user activities under specific trends. This combination helps to capture more complex time-related relationships, thereby improving the prediction accuracy. Although adding the output of the time series model to the input data of the random forest seems to only add data at one time point, in fact, it represents a deeper level of information integration. The updated matrix not only contains current information but also includes trend and expected change information brought by the time series model, which cannot be directly expressed in static features. The output of the time series can make the random forest model more sensitive to upcoming changes, enabling the model to no longer rely solely on past static features but be able to dynamically adjust the prediction strategy, helping the random forest model better capture the complex dependence relationships of time features, thereby improving the overall prediction accuracy and reliability.

[0069] A book order data management method provided by an embodiment of the present invention can more comprehensively capture the characteristics of the order quantity and user activity changing over time by analyzing data in multiple time dimensions, so as to provide more accurate predictions and recommendations in different time periods; by making full use of the time characteristics implicit in historical data and introducing future data predicted by a time series model, it can not only capture the static time characteristics in historical data, but also identify the trend and periodic changes in the time series. The data output by the time series model provides predictions of future order volume and user activity, enabling the random forest model to make more accurate decisions based on this forward-looking information, helping the model better understand the complex dependence relationships of time characteristics. The introduction of dynamic information enables the model to better cope with complex market changes and improves the overall prediction accuracy; the platform can adjust the book recommendation strategy in real time based on the calculated priority parameters, facilitating the platform to formulate a more refined and personalized book recommendation strategy to meet the needs of different time periods and user groups, giving priority to recommending relevant books during the peak period of order volume or user activity, improving the relevance of recommendations and user satisfaction, and thus enhancing the user experience and conversion rate.

[0070] The above are only specific implementation manners of the present invention, enabling those skilled in the art to understand or implement the present invention. Various modifications to these embodiments will be obvious to those skilled in the art. The parts not described in detail in this specification belong to the prior art well known to those skilled in the art.

Claims

1. A method for managing book order data, characterized in that, Including: Collecting the historical book order data of the collection platform, extracting the order quantity feature and order time feature of each type of book from the historical book order data, and constructing multiple book order feature matrices based on the order quantity feature and order time feature of each type of book; Inputting the book order feature matrix into the order time feature analysis model, using the order time feature as the input feature of the order time feature analysis model, and using the order quantity feature as the label to train the order time feature analysis model for each type of book and generate the target order time feature of each type of book; Collecting the platform user data corresponding to the historical book order data, extracting the user quantity feature and user time feature from the platform user data, constructing the platform user feature matrix based on the user quantity feature and user time feature, inputting the platform user feature matrix into the platform user feature analysis model, using the user time feature as the input feature of the platform user feature analysis model, and using the user quantity feature as the label to train the platform user feature analysis model and generate the target user time feature of the platform; Inputting the target user time feature and the target order time feature of each type of book into the order time feature analysis model corresponding to each type of book to generate the order prediction parameters of multiple time features; Inputting the target user time feature and the target order time feature of each type of book into the platform user feature analysis model to generate the user prediction parameters of multiple time features; Extracting the order-user association parameters of each type of book under the target order time feature and target user time feature from the historical book order data and platform user data, calculating the priority parameters of each type of book under multiple time features according to the order prediction parameters, user prediction parameters and order-user association parameters, and adjusting the recommendation strategy of multiple types of books on the platform based on the priority parameters.

2. The method for managing book order data according to claim 1, wherein Extracting the order-user association parameters of each type of book under the target order time feature and target user time feature from the historical book order data and platform user data, including: Extracting the order quantity data of each type of book under each target order time feature and target user time feature from the historical book order data, and extracting the platform active user data under each target order time feature and target user time feature from the platform user data; Calculating the order-user association parameters of each type of book under each target order time feature and target user time feature according to the order quantity data and platform active user data, and the order-user association parameter is the ratio of the order quantity to the platform active users.

3. The method for managing book order data according to claim 1, characterized in that, Calculating the priority parameters of each type of book under multiple time features according to the order prediction parameters, user prediction parameters and order-user association parameters, including: Processing the order prediction parameters and user prediction parameters under each time feature based on the preset time feature unit to obtain the order normalization parameter and user normalization parameter, and determining the recommendation identifier under each time feature according to the order normalization parameter; Associating the order normalization parameters, user normalization parameters, order-user association parameters and recommendation identifiers under multiple time features and constructing an order feature data set; Input the order feature dataset into the associated feature analysis model. Use the order normalization parameter, user normalization parameter, and order-user association parameter in the order feature dataset as the input features of the associated feature analysis model, and use the recommendation identifier as the label. Output the target scores corresponding to the order normalization parameter, user normalization parameter, and order-user association parameter respectively through the associated feature analysis model, and calculate the priority parameters of each type of book under multiple time features based on the target scores; Among them, calculating the priority parameters of each type of book under multiple time features based on the target scores includes determining the weight parameters corresponding to the order normalization parameter, user normalization parameter, and order-user association parameter respectively according to the target scores, and performing weighted calculations on the order normalization parameter, user normalization parameter, and order-user association parameter under each time feature based on the weight parameters to obtain the priority parameters of each type of book under multiple time features.

4. A method for managing book order data according to claim 3, characterized in that, Process the order prediction parameter and user prediction parameter under each time feature based on the preset time feature unit to obtain the order normalization parameter and user normalization parameter, and determine the recommendation identifier under each time feature according to the order normalization parameter, including: Perform normalization processing on each time feature based on the preset time feature unit, calculate the normalization parameter of each time feature, correct the order prediction parameter and user prediction parameter under each time feature respectively according to the normalization parameter to obtain the order normalization parameter and user normalization parameter under each time feature, and determine the recommendation identifier under each time feature based on the preset quantity threshold and the order normalization parameter, including recording the recommendation identifier under the time feature where the order normalization parameter is greater than the preset quantity threshold as 1, otherwise recording it as 0.

5. A method for managing book order data according to claim 1, characterized in that, For the book order feature matrix and the platform user feature matrix, it also includes: Extract the time series data of each type of book order from the historical book order data and construct an order time series dataset. Train an order time series analysis model through the order time series dataset, predict the order time series prediction quantity of each type of book in the first preset period through the order time series analysis model, determine the order time feature corresponding to the first preset period and the order quantity feature corresponding to the order time series prediction quantity in the first preset period, add the order time feature corresponding to the first preset period and the order quantity feature corresponding to the order time series prediction quantity in the first preset period to the book order feature matrix corresponding to the book order of the corresponding type, and train an order time feature analysis model for each type of book through the book order feature matrix; Extract the time series data of the number of users from the platform user data and construct a user time series data set. Train a user time series analysis model through the user time series data set. Predict the user time series prediction quantity in the second preset period through the order time series analysis model. Determine the user time characteristics corresponding to the second preset period and the user quantity characteristics corresponding to the user time series prediction quantity in the second preset period. Add the user time characteristics corresponding to the second preset period and the user quantity characteristics corresponding to the user time series prediction quantity in the second preset period to the platform user feature matrix. Train a platform user feature analysis model based on the platform user feature matrix.

6. The method for managing book order data according to claim 1, characterized in that, Select the order time characteristics with the feature importance scores of multiple features of each type of book greater than the preset scoring threshold as the target order time characteristics of each type of book. Select the user time characteristics with the feature importance scores of multiple features greater than the preset scoring threshold as the target user time characteristics of the platform.

7. A method for managing book order data according to claim 6, characterized in that, The feature analysis model is constructed based on the random forest model, and the time series analysis model is constructed based on the ARIMA model.

Images