Sales trend prediction method and related device

Abstract

The invention discloses a sales trend prediction method and a related device, and the method comprises the steps: obtaining the real-time live broadcast data of a target live broadcast room for a target commodity at a current moment in a live broadcast sales process through the target live broadcast room, and enabling the real-time live broadcast data to be used for reflecting the sales condition of the target commodity under the current interaction data, and to be the latest state information; and dynamic demand and short-term fluctuation of users and markets can be reflected. The historical sales data of the target commodity is obtained, the historical sales data is used for reflecting the sales condition of the target commodity under the historical interaction data, and the historical sales data is the basis of sales prediction and reflects the user and market demands under general conditions, namely, reflects the long-term trend of the user and the market. In this way, the real-time live broadcast data and the historical sales data can be combined, prediction can be performed through the sales trend prediction model, and the sales trend of the target commodity in the target time period is obtained, so that the accuracy and timeliness of sales trend prediction are improved, and the real-time performance and accuracy of sales trend prediction are ensured.

Description

Technical Field

[0001] This application relates to the field of computer technology, and in particular to a sales trend forecasting method and related apparatus. Background Technology

[0002] With the rapid development of internet technology and the widespread adoption of mobile devices, live-streaming e-commerce has emerged as an important form of e-commerce. It's a new shopping model that integrates entertainment, shopping, and social interaction. Through live streaming, it presents product displays and the sales process directly to consumers, enhancing the shopping experience and boosting sales. With continuous technological advancements and evolving consumer demands, the future development prospects of live-streaming e-commerce are even broader.

[0003] However, as the live-streaming e-commerce industry continues to grow, while merchants enjoy the high traffic and conversion rates brought by live-streaming, they also face many challenges. Among these, accurately predicting real-time sales trends during live-streaming has become a critical issue that merchants urgently need to address. Related technologies utilize historical sales data and analyze it through simple statistical methods or moving averages to obtain sales trend predictions.

[0004] However, in the rapidly changing landscape of live-streaming e-commerce, this method struggles to reflect market changes and user demands in real time, resulting in delayed and inaccurate sales trend predictions. Summary of the Invention

[0005] To address the aforementioned technical issues, this application provides a sales trend forecasting method and related apparatus. This method can fully understand the long-term trends of the market and users based on a large amount of historical sales data, and can also combine real-time live data to reflect market changes and user needs in real time, thereby improving the accuracy and timeliness of sales trend forecasting and ensuring its real-time performance and accuracy.

[0006] The embodiments of this application disclose the following technical solutions:

[0007] On one hand, embodiments of this application provide a sales trend forecasting method, the method comprising:

[0008] During the live sales process through the target live room, real-time live data of the target product in the target live room at the current moment is obtained. The real-time live data is used to reflect the sales situation of the target product under the current interaction data. The current interaction data is the interaction data corresponding to the current moment.

[0009] Obtain historical sales data of the target product. The historical sales data is used to reflect the sales situation of the target product under historical interaction data. The historical interaction data is the interaction data corresponding to different times before the current time.

[0010] Based on the real-time live data and the historical sales data, a sales trend prediction model is used to predict the sales trend of the target product within a target time period. The sales trend prediction model is a model trained on a sample dataset for sales prediction, and the target time period is a period of time after the current moment.

[0011] On one hand, embodiments of this application provide a sales trend prediction device, the device comprising an acquisition unit and a prediction unit:

[0012] The acquisition unit is used to acquire real-time live streaming data of the target product in the target live streaming room at the current moment during the live streaming sales process through the target live streaming room. The real-time live streaming data is used to reflect the sales situation of the target product under the current interaction data. The current interaction data is the interaction data corresponding to the current moment.

[0013] The acquisition unit is further configured to acquire historical sales data of the target product, the historical sales data being used to reflect the sales performance of the target product under historical interaction data, and the historical interaction data being the interaction data corresponding to different times before the current time.

[0014] The prediction unit is used to predict the sales trend of the target product within a target time period based on the real-time live data and the historical sales data using a sales trend prediction model. The sales trend prediction model is a model trained on a sample dataset for sales prediction, and the target time period is a period of time after the current moment.

[0015] In one possible implementation, the prediction unit is used for:

[0016] The real-time live data is used to extract features to obtain a first feature, and the historical sales data is used to extract features to obtain a second feature;

[0017] Arrange the first and second features in chronological order according to their corresponding time points to obtain time series feature data;

[0018] Based on the time series feature data, the sales trend prediction model is used to predict the sales trend of the target product within the target time period.

[0019] In one possible implementation, the prediction unit is used for:

[0020] Determine the stationarity of the time series feature data;

[0021] If the stationarity is lower than a preset threshold, the time series feature data is differentially processed by the sales trend prediction model to convert the time series feature data into stationary series feature data.

[0022] The sales trend prediction model is used to predict the stationary sequence feature data to obtain the sales trend of the target product within the target time period.

[0023] In one possible implementation, the prediction unit is used for:

[0024] The sales trend prediction model is updated using the real-time live data to obtain the updated sales trend prediction model;

[0025] Based on the historical sales data, the updated sales trend prediction model is used to predict the sales trend of the target product within the target time period.

[0026] In one possible implementation, the prediction unit is used for:

[0027] Based on the historical sales data and the real-time live data, the sales trend of the target product within the target time period is predicted by the updated sales trend prediction model.

[0028] In one possible implementation, the sales forecasting model includes multiple forecasting sub-models, the forecasting unit being used for:

[0029] Based on the real-time live data and the historical sales data, predictions are made through the multiple prediction sub-models to obtain the prediction results output by each prediction sub-model.

[0030] Obtain the weight of each prediction sub-model in the plurality of prediction sub-models;

[0031] Based on the weights of each prediction sub-model and the corresponding prediction results, the sales trend of the target product within the target time period is determined.

[0032] In one possible implementation, the device further includes a training unit:

[0033] The training unit is used to acquire sample live streaming data of the sample product in the historical live streaming room at a historical moment during the process of live streaming sales through the historical live streaming room.

[0034] Obtain historical sample sales data for the sample product, which reflects the sales situation of the sample product at different times before the historical moment.

[0035] Based on the sample live data and the historical sample sales data, an initial prediction model is used to predict the sales trend of the sample product within a sample time period, where the sample time period is a period of time after the historical moment.

[0036] Based on the difference between the predicted sales trend and the actual sales trend, the model parameters of the initial prediction model are adjusted to obtain the sales trend prediction model.

[0037] In one possible implementation, the training unit is used for:

[0038] Based on the difference between the predicted sales trend and the actual sales trend, the initial prediction model is trained to obtain an intermediate prediction model;

[0039] The performance of the intermediate prediction model is evaluated using the objective evaluation function to obtain the evaluation results;

[0040] The model parameters of the intermediate prediction model are optimized based on the evaluation results to obtain the sales trend prediction model.

[0041] In one possible implementation, the apparatus further includes a generation unit:

[0042] The acquisition unit is also used to acquire the inventory data at the current moment;

[0043] The generation unit is used to generate an inventory scheduling strategy based on the sales trend of the target product within a target time period and the inventory data.

[0044] In one possible implementation, the generating unit is used for:

[0045] Based on the sales trend of the target product within the target time period and the inventory data, the inventory identification result is determined;

[0046] If the inventory identification result is the target result, the inventory scheduling strategy is generated by combining the factors affecting inventory scheduling.

[0047] In one possible implementation, the device further includes an early warning unit:

[0048] The early warning unit is used to send an early warning notification if it determines that the sales trend of the target product and the inventory data meet the early warning rules within the target time period.

[0049] On one hand, embodiments of this application provide a computer device, the computer device including a processor and a memory:

[0050] The memory is used to store computer programs and to transfer the computer programs to the processor;

[0051] The processor is configured to execute the method described in any of the foregoing aspects according to instructions in the computer program.

[0052] In one aspect, embodiments of this application provide a computer-readable storage medium for storing a computer program that, when executed by a processor, causes the processor to perform the methods described in any of the foregoing aspects.

[0053] On one hand, embodiments of this application provide a computer program product, including a computer program that, when executed by a processor, implements the method described in any of the foregoing aspects.

[0054] As can be seen from the above technical solution, for live-streaming e-commerce scenarios, this application provides a method for real-time prediction of product sales trends over a future period during live-streaming. For any moment in the live-streaming process, such as the current moment, during the live-streaming sales process through the target live-streaming room, real-time live-streaming data for the target product in the target live-streaming room at the current moment is obtained. This real-time live-streaming data reflects the sales situation of the target product under the current interaction data. The current interaction data is the interaction data corresponding to the current moment; therefore, the real-time live-streaming data is the latest status information, reflecting the dynamics and short-term fluctuations of user and market demand. Additionally, historical sales data for the target product can also be obtained. This historical sales data reflects the sales situation of the target product under historical interaction data. The historical interaction data is the interaction data corresponding to different moments before the current moment. Historical sales data is the basis for sales prediction, reflecting user and market demand under general circumstances, i.e., reflecting the long-term trends of users and the market. Thus, by combining real-time live-streaming data and historical sales data, a sales trend prediction model can be used to predict the sales trend of the target product within a target time period. The sales trend prediction model is a model trained on a sample dataset for sales prediction, and the target time period is a period after the current moment. This allows for a comprehensive understanding of long-term market and user trends based on a large amount of historical sales data, while also incorporating real-time live data to reflect market changes and user needs in real time. This improves the accuracy and timeliness of sales trend forecasting, ensuring its real-time nature and accuracy. Attached Figure Description

[0055] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0056] Figure 1 An application scenario architecture diagram of a sales trend forecasting method provided in this application embodiment;

[0057] Figure 2 A flowchart illustrating a sales trend forecasting method provided in this application embodiment;

[0058] Figure 3 An example diagram of an interface for live sales in a target live streaming room, provided as an embodiment of this application;

[0059] Figure 4 A flowchart illustrating a training method for a sales trend prediction model provided in this application embodiment;

[0060] Figure 5 A structural diagram of a sales trend forecasting device provided in an embodiment of this application;

[0061] Figure 6 A structural diagram of a terminal provided in an embodiment of this application;

[0062] Figure 7 This is a structural diagram of a server provided in an embodiment of this application. Detailed Implementation

[0063] The embodiments of this application will now be described with reference to the accompanying drawings.

[0064] In the live-streaming e-commerce industry, common sales trend predictions are mainly based on simple prediction models using historical sales data. This involves analyzing historical sales data using simple statistical methods or moving averages to obtain sales trend predictions.

[0065] However, this approach has the following technical problems: 1. Lack of real-time performance: the relevant technologies lag in data collection and processing, failing to reflect market changes and user needs in a timely manner. 2. Insufficient accuracy: the simple models used in the relevant technologies lack the ability to accurately predict sales trends, leading to unstable prediction results.

[0066] To address the aforementioned technical problems, this application provides a sales trend prediction method. This method is a real-time prediction of a product's sales trend over a future period. For any moment during a live broadcast, such as the current moment, it combines real-time live broadcast data and historical sales data for the target product at that moment, and uses a sales trend prediction model to predict the target product's sales trend within the target time period. This approach not only provides a comprehensive understanding of long-term market and user trends based on extensive historical sales data, but also reflects market changes and user needs in real time by incorporating real-time live broadcast data, thus improving the accuracy and timeliness of sales trend prediction and ensuring its real-time performance and accuracy.

[0067] The sales trend forecasting method provided in this application can be applied to computer equipment with sales trend forecasting capabilities, such as terminal equipment and servers.

[0068] Specifically, terminal devices can be desktop computers, laptops, smartphones, tablets, IoT devices, and portable wearable devices. IoT devices can be smart speakers, smart TVs, smart air conditioners, smart in-vehicle devices, etc. Smart in-vehicle devices can be in-vehicle navigation terminals and in-vehicle computers, etc. Portable wearable devices can be smartwatches, smart bracelets, head-mounted devices, etc., but are not limited to these.

[0069] The server can be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server or server cluster that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms. Terminal devices and servers can be connected directly or indirectly via wired or wireless communication; this application does not impose any restrictions on this.

[0070] To facilitate understanding of the sales trend prediction method provided in the embodiments of this application, the following example illustrates the application scenario of the sales trend prediction method, with the method being executed by a server.

[0071] like Figure 1 As shown, Figure 1 An application scenario architecture diagram of a sales trend forecasting method is shown. This application scenario may include a terminal 101 and a server 102, which can communicate via a communication network. The communication network uses standard communication technologies and / or protocols, typically the Internet, but can also be any network, including but not limited to Bluetooth, local area network (LAN), metropolitan area network (MAN), wide area network (WAN), mobile, private network, or any combination of virtual private network. In some embodiments, customized or dedicated data communication technologies may be used to replace or supplement the aforementioned data communication technologies.

[0072] A live-streaming sales platform can run on terminal 101, and server 102 can provide live-streaming sales services for the platform. Hosts can conduct live-streaming e-commerce through terminal 101; that is, the live stream enters the target live-streaming room through terminal 101, and then conducts live-streaming sales within that room. The target live-streaming room can be any room specifically designed for live-streaming sales. In the live-streaming e-commerce scenario, the host can be anyone who introduces products to users and directly conducts sales online. This could be the merchant themselves, or a well-known online influencer or celebrity who bears advertising responsibility.

[0073] During the live stream, users can act as viewers and enter the target live stream room through the corresponding terminal 101 to watch the live stream. The host can interact with users in the target live stream room, such as introducing products, chatting, answering questions, and giving away benefits, etc. Users can purchase products they are interested in within the target live stream room.

[0074] For any point in the live stream, such as the current moment, during the live sales process through the target live stream room, server 102 can obtain the real-time live stream data of the target product in the target live stream room at the current moment. The target product can be any product sold in the target live stream room. The real-time live stream data is used to reflect the sales status of the target product under the current interaction data. The current interaction data is the interaction data corresponding to the current moment. Therefore, the real-time live stream data is the latest status information and can reflect the dynamics and short-term fluctuations of user and market demand.

[0075] In addition, server 102 can also obtain historical sales data of the target product. Historical sales data is used to reflect the sales situation of the target product under historical interaction data. Historical interaction data is the interaction data corresponding to different times before the current time. Historical sales data is the basis for sales forecasting and reflects the user and market demand under normal circumstances, that is, it reflects the long-term trend of users and the market.

[0076] In this way, when making sales trend predictions, server 102 can combine real-time live data and historical sales data to predict the sales trend of the target product within a target time period using a sales trend prediction model. This model, trained on a sample dataset, is used for sales forecasting, and the target time period is a period following the current moment. This approach allows for a comprehensive understanding of long-term market and user trends based on extensive historical sales data, while also incorporating real-time live data to reflect market changes and user needs in real time. This improves the accuracy and timeliness of sales trend predictions, ensuring their real-time nature and accuracy.

[0077] It should be noted that after receiving the sales trend, server 102 can send the sales trend to the merchant's terminal so that the merchant can understand the sales trend in real time. Figure 1 Taking the server-executed sales trend prediction method as an example, in some cases, the terminal can also have similar functions to the server, thereby executing the sales trend prediction method by the terminal. This terminal could be, for example, a merchant's terminal, which could obtain historical sales data and real-time live streaming data from the server, and then combine the historical sales data and real-time live streaming data to predict the sales trend. Alternatively, the terminal and server can jointly execute the sales trend prediction method provided in this application embodiment; this embodiment does not limit this approach.

[0078] The sales trend prediction method provided in this application can be applied to various live-streaming e-commerce scenarios. During the live-streaming sales process, sales trends can be predicted in real time, thereby understanding the sales trends for a period of time in the future. This allows for timely adjustments to sales strategies during the live-streaming sales process, such as interaction methods, and can even provide important basis for inventory management and scheduling.

[0079] It should be noted that in the specific implementation of this application, the entire process may involve user information and other related data. When the above embodiments of this application are applied to specific products or technologies, separate consent or permission from the user is required, and the collection, use and processing of related data must comply with the relevant laws, regulations and standards of the relevant countries and regions.

[0080] The following describes in detail a sales trend forecasting method provided in this application through method embodiments.

[0081] See Figure 2 , Figure 2 A flowchart of a sales trend forecasting method is shown, which may include steps S201-S203, as detailed below:

[0082] S201. During the live sales process through the target live room, obtain the real-time live data of the target product in the target live room at the current moment.

[0083] In live-streaming e-commerce, hosts can enter a target live-streaming room via a terminal to conduct live sales, and users can enter as viewers to watch. During the live stream, the host introduces and sells various products in the target live-streaming room to encourage viewers to make purchases. See also Figure 3 As shown, Figure 3 This is an example screenshot of an interface for live-streaming sales in a target live-streaming room. Taking food as an example, the host introduces the product in the target live-streaming room, and users watch and purchase items they are interested in.

[0084] The sales trend of each product in the target live streaming room directly affects product scheduling, inventory management, and strategy formulation. Therefore, real-time sales trend prediction is crucial. Real-time sales trend prediction refers to predicting the sales trend of a product in real time over a period of time. In this embodiment, to improve the real-time performance and accuracy of sales trend prediction, real-time sales trend prediction mainly refers to combining real-time live streaming data and historical sales data to predict the sales trend of a product in real time over a period of time.

[0085] Real-time live stream data reflects the sales performance of the target product at the current moment, where interactive data refers to the data generated at that specific instant. Interactive data can refer to the interactions between the streamer and users in the target live stream room, such as product introductions, live chat, answering questions, distributing giveaways, and even returns and refunds. Sales performance can include metrics such as sales revenue, order quantity, time, and product attributes.

[0086] It is understood that in this embodiment, real-time data acquisition can be used to obtain real-time live streaming data from the live streaming sales platform, that is, the live streaming sales platform is used as the data source and connected to the computer device for sales trend prediction. Of course, if there are other data sources that can provide real-time live streaming data, they can also be connected to the computer device for sales trend prediction. Each data source can be represented by Di, and the set of data sources connected to the computer device for sales trend prediction can be represented by D, then D = {D1, D2, ..., Dn}, i = 1, 2, ..., n.

[0087] The process of obtaining real-time live data from a data source can be carried out through polling, push notifications, or other methods. Assume that each data source Di generates new data periodically, and let di(t) represent the real-time live data obtained from data source Di at time t.

[0088] In some cases, the collected live stream data may contain noise, missing values, or other issues, requiring real-time processing and cleaning. In such situations, obtaining the real-time live stream data for the target product in the target live stream room at the current moment can be achieved by collecting initial live stream data, processing and cleaning it to obtain the real-time live stream data. One possible implementation is to use C(di(t)) to represent the cleaning operation on the initial live stream data di(t), and then use the cleaned data ci(t) as the real-time live stream data.

[0089] In this embodiment, high-efficiency real-time data processing components, including Apache Pulsar and Apache Flink, are used for real-time data acquisition, processing, and analysis.

[0090] Apache Pulsar is a distributed messaging and stream processing platform characterized by high throughput, low latency, and horizontal scalability. In this embodiment, Pulsar's message queue functionality is utilized to achieve real-time data acquisition and transmission. Real-time data, such as live streaming data and inventory data, are sent to message topics via Pulsar producers. These topics are then subscribed to in real-time by Pulsar consumers, and the data is transmitted to subsequent processing flows.

[0091] Apache Flink is an open-source stream processing framework that supports high-throughput, low-latency real-time data processing. In this embodiment, Flink is used to process and analyze initial live stream data. Flink's streaming capabilities enable real-time cleaning, transformation, and aggregation of the initial live stream data. The initial live stream data is transmitted to Flink via Pulsar, where Flink processes the data according to pre-designed real-time processing logic and outputs the results to downstream systems for further analysis and application.

[0092] By combining Apache Pulsar and Apache Flink, efficient real-time data acquisition and processing can be achieved, ensuring timely acquisition and processing of sales and inventory data from live-streaming e-commerce merchants, providing reliable data support for subsequent sales trend prediction and inventory management.

[0093] S202. Obtain the historical sales data of the target product.

[0094] This application embodiment can also obtain historical sales data of the target product. Historical sales data reflects the sales performance of the target product under historical interaction data, which refers to interaction data from different times prior to the current moment. Sales performance data may include, for example, sales amount, order quantity, time, and product attributes. Historical sales data can be represented by the symbol "hi".

[0095] It should be noted that the execution order of S201 and S202 is not limited in the embodiments of this application. S201 can be executed first and then S202 can be executed, or S202 can be executed first and then S201 can be executed.

[0096] S203. Based on the real-time live data and the historical sales data, a sales trend prediction model is used to predict the sales trend of the target product within the target time period.

[0097] By obtaining real-time live data and historical sales data, a sales trend prediction model can be used to predict the sales trend of the target product within a target time period. The target time period is a period of time following the current moment.

[0098] The sales trend prediction algorithm is one of the key steps in realizing the core function of sales trend prediction. This algorithm utilizes historical and real-time sales data to predict future sales trends through analysis and modeling. In this embodiment, the sales trend prediction model is a model trained on a sample dataset for sales forecasting. This embodiment does not limit the model structure of the sales trend prediction model; a suitable model structure can be selected for training. Common sales trend prediction models include time series analysis models, neural network models, deep learning models, or multiple basic prediction sub-models. Time series analysis models include the Autoregressive Integrated Moving Average (ARIMA) model and the Seasonal Autoregressive Integrated Moving Average (SARIMA) model. The SARIMA model is a time series prediction model, particularly suitable for data with seasonal variations. It is an extension of the ARIMA model that incorporates seasonal factors. Neural network models can be, for example, Long Short-Term Memory (LSTM) networks or gated recurrent units (GRUs). Deep learning models can be, for example, recurrent neural networks (RNNs).

[0099] After obtaining historical sales data, it can be preprocessed, including data cleaning and outlier handling. Therefore, in one possible implementation, S203 could involve preprocessing the historical sales data to obtain preprocessed data. Based on the real-time live data and the preprocessed data, a sales trend prediction model can be used to predict the sales trend of the target product within a target time period. Here, P(hi) can represent the preprocessing operation on the historical sales data hi, and the processed data can be represented as pi.

[0100] By forecasting sales trends, businesses can gain timely insights into future sales performance. This enables them to develop more accurate marketing strategies, adjust inventory management, and improve sales efficiency and profitability.

[0101] As can be seen from the above technical solution, for live-streaming e-commerce scenarios, this application provides a method for real-time prediction of product sales trends over a future period during live-streaming. For any moment in the live-streaming process, such as the current moment, during the live-streaming sales process through the target live-streaming room, real-time live-streaming data for the target product in the target live-streaming room at the current moment is obtained. This real-time live-streaming data reflects the sales situation of the target product under the current interaction data. The current interaction data is the interaction data corresponding to the current moment; therefore, the real-time live-streaming data is the latest status information, reflecting the dynamics and short-term fluctuations of user and market demand. Additionally, historical sales data for the target product can also be obtained. This historical sales data reflects the sales situation of the target product under historical interaction data. The historical interaction data is the interaction data corresponding to different moments before the current moment. Historical sales data is the basis for sales prediction, reflecting user and market demand under general circumstances, i.e., reflecting the long-term trends of users and the market. Thus, by combining real-time live-streaming data and historical sales data, a sales trend prediction model can be used to predict the sales trend of the target product within a target time period. The sales trend prediction model is a model trained on a sample dataset for sales prediction, and the target time period is a period after the current moment. This allows for a comprehensive understanding of long-term market and user trends based on a large amount of historical sales data, while also incorporating real-time live data to reflect market changes and user needs in real time. This improves the accuracy and timeliness of sales trend forecasting, ensuring its real-time nature and accuracy.

[0102] It should be noted that, compared with related technologies, the embodiments of this application mainly introduce real-time live streaming data during the live streaming process, and combine the real-time live streaming data with historical sales data to make predictions and obtain the sales trend of the target product within a target time period. The real-time live streaming data introduced can play different roles in sales trend prediction. In one possible implementation, the introduced real-time live streaming data mainly expands the dataset on which the prediction is based based on historical sales data, thereby improving the accuracy of the prediction.

[0103] Based on this, in S203, a sales trend prediction model is used to predict the sales trend of the target product within a target time period based on real-time live streaming data and historical sales data. This can be achieved by extracting features from the real-time live streaming data to obtain the first feature, and from the historical sales data to obtain the second feature. In other words, both real-time live streaming data and historical sales data are used as the dataset for prediction, and features are extracted from each data point. Then, the first and second features are arranged in chronological order according to their corresponding time points to obtain time-series feature data. Since historical sales data may include multiple data points, the second feature may include multiple features, each corresponding to a different time point. Therefore, all features are arranged in chronological order to obtain time-series feature data. The time-series feature data, denoted as yt, can be represented as time t, where t = 1, 2, ..., and yt represents the observed value at time t. Finally, based on the time-series feature data, a sales trend prediction model is used to predict the sales trend of the target product within the target time period.

[0104] It's understandable that sales data can include metrics such as sales revenue, order quantity, time frame, and product attributes. Therefore, when extracting features, appropriate features can be selected for modeling based on the characteristics of historical sales data or real-time live data and business needs. Possible features include sales revenue, order quantity, time frame, and product attributes. Assuming m features are selected, let F = {f1, f2, ..., f...} m Let} represent the feature set. Features at different times can be represented as Ft. Arranging the features at different times in chronological order yields the time series feature data yt.

[0105] In this application embodiment, when making sales trend predictions, the dataset on which the prediction is based is expanded by using real-time live data on the basis of historical sales data. The prediction is made on a broader dataset that integrates historical sales data and real-time live data. It not only considers the long-term trends and periodic changes reflected by historical sales data, but also incorporates the latest dynamic information reflected by real-time live data, thereby improving the accuracy of the prediction.

[0106] It should be noted that, in this case, the sales trend forecasting model can be a time series analysis model. This application primarily uses SARIMA as an example of a time series analysis model. The general form of the SARIMA model can be expressed as:

[0107]

[0108] Where, φ p (B) and θ q(B) is the polynomial of the autoregressive and moving average components, Φ P (B s ) and Θ Q (B s ) is a polynomial consisting of the seasonal autoregressive component and the seasonal moving average component, Δ d and Represent the difference operations for non-seasonal and seasonal operations, respectively, ∈ t It is a white noise sequence.

[0109] As can be seen from the general form of the time series analysis model described above, this sales trend prediction model includes differencing operations, which are an important means of transforming non-stationary time series into stationary time series. Therefore, when using the sales trend prediction model for forecasting, it is necessary to first determine whether the time series feature data is a stationary time series. If not, differencing operations are required to transform it into a stationary time series before proceeding with subsequent operations. In this case, based on the time series feature data, the sales trend prediction model can be used to predict the sales trend of the target product within the target time period by determining the stationarity of the time series feature data. The level of stationarity reflects whether the time series feature data is a stationary time series; the higher the stationarity, the more likely the time series feature data is to be a stationary time series, and vice versa. If the stationarity is below a preset threshold, it indicates that the time series feature data is more likely to be a non-stationary time series. Therefore, the sales trend prediction model can be used to perform differencing operations on the time series feature data to transform it into stationary series feature data. The stationary series feature data is a stationary time series, and then the sales trend prediction model can be used to predict the sales trend of the target product within the target time period.

[0110] In forecasting, transforming non-stationary time series into stationary ones eliminates time dependencies in the time series data, making the data more stable. It also removes common error interference, enhancing robustness against disturbances. This makes the data after differencing clearer and more accurate, which is beneficial for subsequent sales trend forecasting.

[0111] Understandably, besides using SARIMA as a sales trend forecasting model through the methods described above, deep learning models such as RNN, LSTM, or GRU can also be considered for sales trend forecasting. These sales trend forecasting models are better able to capture the non-linear relationships and long-term dependencies of time series data, thus ensuring the accuracy of sales trend forecasting.

[0112] In another possible implementation, the real-time live data is mainly used to adjust the sales trend prediction model so that it can capture short-term fluctuations and thus improve prediction accuracy.

[0113] In this scenario, S203 uses a sales trend prediction model based on real-time live streaming data and historical sales data to predict the sales trend of the target product within the target time period. This can be achieved by first updating the sales trend prediction model using real-time live streaming data. The updated model, trained on real-time live streaming data, learns the latest dynamic information reflected in the data and captures short-term fluctuations, making it more suitable for predicting sales trends at the current moment. Then, based on historical sales data, the updated model is used to predict the sales trend of the target product within the target time period.

[0114] Real-time live data may include current sales figures, market feedback, and the effectiveness of promotional activities. This data is crucial for adjusting forecasting models and capturing short-term fluctuations. Therefore, when forecasting, one can first use real-time live data to update the sales trend forecasting model, and then use the updated model to make predictions, thereby improving the accuracy of sales trend forecasting by improving the accuracy of the sales trend forecasting model.

[0115] When predicting the sales trend of a target product within a target time period using an updated sales trend prediction model based on historical sales data, real-time live streaming data can be further incorporated. This allows for predictions based on both historical and live streaming data, using a broader dataset that integrates historical and live streaming data. This approach considers not only the long-term trends and cyclical changes reflected in historical sales data but also incorporates the latest dynamic information from live streaming data, thereby further improving prediction accuracy.

[0116] In one possible implementation, embodiments of this application may also employ an ensemble learning method for sales trend prediction. The ensemble learning method may be such as Random Forest or Gradient Boosting Tree, thereby combining multiple basic prediction sub-models to predict sales trends, thereby improving prediction accuracy and stability.

[0117] In this scenario, the sales forecasting model comprises multiple forecasting sub-models. Based on real-time live streaming data and historical sales data, a sales trend forecasting model is used to predict the sales trend of the target product within a target time period. This can be achieved by using multiple forecasting sub-models to perform predictions separately, obtaining the prediction results output by each sub-model. The weights of each forecasting sub-model are then determined, and based on these weights and their corresponding prediction results, the sales trend of the target product within the target time period is identified.

[0118] Different prediction sub-models may use different ensemble learning methods, and different ensemble learning methods have different prediction accuracies. Therefore, when setting the weights of prediction sub-models, the higher the prediction accuracy of a prediction sub-model, the greater its weight.

[0119] The embodiments of this application combine multiple basic prediction sub-models to predict sales trends, thereby improving prediction accuracy and stability.

[0120] Compared with related technologies, another important difference in the embodiments of this application is that advanced machine learning algorithms are used to train a sales trend prediction model with better sales trend prediction performance, thereby achieving greater accuracy and stability in sales trend prediction.

[0121] The training process for the sales trend prediction model is as follows: During live sales through historical live streaming rooms, sample live streaming data for sample products at historical moments are obtained (see...). Figure 4 (As shown in S401); Obtain historical sample sales data for the sample products. This historical sample sales data reflects the sales performance of the sample products at different times prior to the historical point in time (see...). Figure 4 As shown in S402); based on sample live data and historical sample sales data, an initial prediction model is used to predict the sales trend of the sample product within a sample time period, which is a period of time after the historical time (see...). Figure 4 As shown in S403; based on the difference between the predicted sales trend and the actual sales trend, the model parameters of the initial prediction model are adjusted to obtain the sales trend prediction model (see S403). Figure 4 (As shown in S404).

[0122] Taking SARIMA, a sales trend forecasting model, as an example, the derivation process of the SARIMA model includes the following steps:

[0123] Determine the model order: First, it is necessary to determine the order p, d, q, P, D, Q, s of the sales trend forecasting model. This can be done by observing the graphs of the autocorrelation function (ACF) and the partial autocorrelation function (PACF).

[0124] Difference operation: If the time series feature data obtained by feature extraction from live sample data and historical sample sales data is not stationary, a difference operation is needed to transform it into a stationary time series. Δ d and These represent difference operations for non-seasonal and seasonal operations, respectively.

[0125] Model Fitting: During training, the parameters of the sales trend prediction model are obtained by fitting the model using methods such as maximum likelihood estimation. The parameters of the sales trend prediction model include the coefficients φ1, φ2, ..., φ of the autoregressive component. p The coefficients θ1, θ2, ..., θ of the moving average component q The coefficients Φ1,Φ2,…,Φ of the seasonal autoregressive component P The coefficients Θ1,Θ2,…,Θ of the seasonal moving average component Q .

[0126] Model Diagnosis: Diagnose and test the fitted sales trend prediction model, including testing the autocorrelation and partial autocorrelation of the residual sequence and whether it conforms to the characteristics of white noise.

[0127] During the training process of the sales trend prediction model, model evaluation and optimization can also be performed. In this case, based on the difference between the predicted sales trend and the actual sales trend, the model parameters of the initial prediction model are adjusted to obtain the sales trend prediction model. One approach is to train the initial prediction model based on the difference between the predicted and actual sales trends to obtain an intermediate prediction model. Then, the performance of the intermediate prediction model is evaluated using a target evaluation function, and the evaluation results reflect the prediction accuracy and stability of the intermediate prediction model. Therefore, the model parameters of the intermediate prediction model can be optimized based on the evaluation results to obtain the final sales trend prediction model.

[0128] During the optimization process, the intermediate prediction model can be represented by M, and the objective evaluation function can be represented by E(M). The model parameters of the intermediate prediction model are optimized by minimizing E(M). This application does not limit the evaluation algorithm; for example, the evaluation algorithm can be cross-validation, etc.

[0129] This application embodiment improves the prediction accuracy and stability of the sales trend prediction model by evaluating the model, tuning the model parameters, and optimizing the model.

[0130] Sales trend forecasting not only helps merchants understand sales trends in a timely manner but also facilitates inventory management. Inventory management can be achieved through an inventory management system, which is a system for managing and controlling merchandise inventory, including functions such as inventory procurement, receiving, issuing, and inventory counting. However, existing inventory management systems rely on human experience and static rules, failing to dynamically adjust based on real-time sales data. This can easily lead to excessive or insufficient inventory, resulting in low inventory utilization. This application's embodiments design an intelligent inventory management algorithm to achieve dynamic scheduling and optimization of inventory, thereby improving inventory management efficiency.

[0131] Intelligent inventory management and scheduling is a key step in optimizing system inventory. Specifically, the server can acquire current inventory data and then generate inventory scheduling strategies based on the sales trends and inventory data of target products within a target time period. The current inventory data can be obtained through real-time monitoring of inventory status and may include indicators such as inventory level, inventory turnover rate, and slow-moving items. Inventory scheduling strategies can refer to strategies that dynamically adjust inventory procurement plans and storage locations, including automatic replenishment, inventory transfer, and promotional activities.

[0132] This application's embodiments utilize predicted sales trends and real-time inventory data to dynamically adjust inventory procurement plans and storage locations, intelligently optimizing inventory management and scheduling. This helps merchants reduce excessive or insufficient inventory, lower inventory costs, increase inventory turnover and capital utilization, and improve inventory management efficiency.

[0133] It should be noted that when generating inventory scheduling strategies, the predicted sales trends and inventory data can provide a general understanding of whether the inventory is sufficient or excessive. In this case, whether to adopt automatic replenishment or inventory transfer requires consideration of other inventory scheduling factors, so as to formulate the inventory scheduling strategy that is most beneficial to the merchant.

[0134] Therefore, based on the sales trends and inventory data of the target product within the target time period, an inventory scheduling strategy can be generated by determining the inventory identification result. The inventory identification result can reflect whether there are problems and potential risks in the inventory. If the inventory identification result is the target result, it indicates that there are problems and potential risks in the inventory. At this time, an inventory scheduling strategy can be generated by combining the factors influencing inventory scheduling.

[0135] This application does not limit the factors affecting inventory scheduling. These factors may include, for example, inventory costs, transfer costs, promotional costs, and supply chain conditions.

[0136] For example, if based on predicted sales trends and inventory data, it is determined that there are problems and potential risks in the inventory, such as the risk of excessive inventory, then inventory transfer may be considered. In order to formulate an inventory scheduling strategy that is more beneficial to the merchant, the transfer cost and the cost of excessive inventory will be comprehensively considered. If the inventory cost is significantly higher than the transfer cost, that is, if inventory transfer is more beneficial to the merchant, then an inventory scheduling strategy can be generated as inventory transfer.

[0137] It should be noted that the embodiments of this application do not limit the method for generating inventory scheduling strategies. In one possible implementation, the inventory scheduling strategy can be generated based on deep reinforcement learning. That is, deep reinforcement learning can be used to optimize inventory management strategies, and more flexible and intelligent inventory management decisions can be achieved through the interaction and learning between the agent and the environment. In another possible implementation, the inventory scheduling strategy can be generated based on the optimization of genetic algorithms. That is, optimization algorithms such as genetic algorithms can be used to explore the optimal inventory scheduling strategy and find the optimal solution by simulating the evolutionary process.

[0138] Based on the predicted sales trends and inventory data, this application identifies whether there are problems and potential risks in the inventory. When there are problems and potential risks in the inventory, it generates an inventory scheduling strategy by combining the factors affecting inventory scheduling, thereby intelligently optimizing inventory management and scheduling and generating an inventory scheduling strategy that is more beneficial to merchants.

[0139] After obtaining the inventory scheduling strategy, inventory optimization can be implemented, that is, the inventory can be optimized and scheduled based on the designed inventory scheduling strategy. The embodiments of this application can automatically execute the inventory scheduling strategy, or can be provided to merchants for reference and manual operation.

[0140] After inventory optimization is implemented, its effectiveness can be evaluated and feedback provided. This involves periodically assessing the optimization results and making adjustments based on the assessment. Continuously optimizing inventory scheduling strategies improves the efficiency and effectiveness of inventory management. This application's embodiments improve the stability, reliability, and usability of the inventory management system by continuously optimizing its performance and functionality. The inventory management system can adapt to the needs of merchants of different sizes and complexities, ensuring its continuous and stable operation and meeting the ever-growing business demands of merchants.

[0141] It should be understood that, after obtaining the predicted sales trends and inventory data, this application embodiment can also issue early warnings to merchants based on the predicted sales trends and inventory data, so as to remind merchants to identify and resolve problems in a timely manner. The early warning method can be to send an early warning notification if it is determined that the sales trend and inventory data of the target product within a target time period meet the early warning conditions.

[0142] This application does not limit the method for determining whether the sales trend and inventory data of a target product meet the early warning conditions within a target time period. One possible implementation involves designing early warning rules, and determining whether the early warning conditions are met by judging whether the sales trend and inventory data conform to these rules. If the sales trend and inventory data conform to the early warning rules, then the early warning conditions are determined to be met. These early warning rules could include situations such as a sales decline exceeding a threshold or inventory levels falling below a safe value. The definition of the early warning rules needs to consider business requirements and the importance of the early warning.

[0143] In another possible implementation, machine learning algorithms can be used to build a more intelligent and accurate early warning system. These algorithms identify and predict potential operational risks. If a potential operational risk is identified based on sales trends and inventory data, then the system determines that the sales trends and inventory data meet the early warning criteria. Machine learning algorithms include anomaly detection algorithms and classification algorithms.

[0144] In another possible implementation, an integrated multi-model approach can be adopted, combining multiple forecasting algorithms and decision support models to provide early warning and decision support from different perspectives, thereby improving the reliability and stability of the inventory management system.

[0145] When sending alert notifications, various methods can be used to send them to merchants, such as email, SMS, and application (APP) push notifications.

[0146] While sending early warning notifications, this embodiment of the application can also send inventory scheduling strategies generated based on predicted sales trends and inventory data to merchants, providing them with intelligent decision support. The inventory management system can provide optimization suggestions and decision-making solutions based on the early warning notifications, helping merchants make more accurate and effective business decisions.

[0147] The embodiments of this application can also periodically evaluate the effectiveness of the early warning and decision support system and optimize it based on the evaluation results. Continuously improving the early warning rules and decision support algorithms enhances the accuracy and practicality of the early warning and decision support system.

[0148] In response to sales anomalies or inventory risks, this application's embodiments can issue timely early warning notifications, providing intelligent decision support for merchants. Merchants can adjust their business strategies promptly based on the early warning notifications and optimization suggestions (such as inventory scheduling strategies) to reduce losses and risks.

[0149] This application's embodiments digitize and intelligently manage the business operations, transforming the traditional experience- and intuition-based management model into a data-driven decision-making model. This enables businesses to make more scientific business decisions, improving the accuracy and efficiency of those decisions.

[0150] It should be noted that, based on the implementation methods provided in the above aspects, this application can be further combined to provide more implementation methods.

[0151] Based on the sales trend forecasting method provided in the foregoing embodiments, this application also provides a sales trend forecasting device 500. See also... Figure 5 As shown, the sales trend prediction device 500 includes an acquisition unit 501 and a prediction unit 502:

[0152] The acquisition unit 501 is used to acquire real-time live streaming data of the target product in the target live streaming room at the current moment during the live streaming sales process through the target live streaming room. The real-time live streaming data is used to reflect the sales situation of the target product under the current interaction data. The current interaction data is the interaction data corresponding to the current moment.

[0153] The acquisition unit 501 is further configured to acquire historical sales data of the target product, wherein the historical sales data is used to reflect the sales situation of the target product under historical interaction data, and the historical interaction data is the interaction data corresponding to different times before the current time.

[0154] The prediction unit 502 is used to predict the sales trend of the target product within a target time period based on the real-time live data and the historical sales data using a sales trend prediction model. The sales trend prediction model is a model trained on a sample dataset for sales prediction, and the target time period is a period of time after the current moment.

[0155] In one possible implementation, the prediction unit is used for:

[0156] The real-time live data is used to extract features to obtain a first feature, and the historical sales data is used to extract features to obtain a second feature;

[0157] Arrange the first and second features in chronological order according to their corresponding time points to obtain time series feature data;

[0158] Based on the time series feature data, the sales trend prediction model is used to predict the sales trend of the target product within the target time period.

[0159] In one possible implementation, the prediction unit is used for:

[0160] Determine the stationarity of the time series feature data;

[0161] If the stationarity is lower than a preset threshold, the time series feature data is differentially processed by the sales trend prediction model to convert the time series feature data into stationary series feature data.

[0162] The sales trend prediction model is used to predict the stationary sequence feature data to obtain the sales trend of the target product within the target time period.

[0163] In one possible implementation, the prediction unit is used for:

[0164] The sales trend prediction model is updated using the real-time live data to obtain the updated sales trend prediction model;

[0165] Based on the historical sales data, the updated sales trend prediction model is used to predict the sales trend of the target product within the target time period.

[0166] In one possible implementation, the prediction unit is used for:

[0167] Based on the historical sales data and the real-time live data, the sales trend of the target product within the target time period is predicted by the updated sales trend prediction model.

[0168] In one possible implementation, the sales forecasting model includes multiple forecasting sub-models, the forecasting unit being used for:

[0169] Based on the real-time live data and the historical sales data, predictions are made through the multiple prediction sub-models to obtain the prediction results output by each prediction sub-model.

[0170] Obtain the weight of each prediction sub-model in the plurality of prediction sub-models;

[0171] Based on the weights of each prediction sub-model and the corresponding prediction results, the sales trend of the target product within the target time period is determined.

[0172] In one possible implementation, the device further includes a training unit:

[0173] The training unit is used to acquire sample live streaming data of the sample product in the historical live streaming room at a historical moment during the process of live streaming sales through the historical live streaming room.

[0174] Obtain historical sample sales data for the sample product, which reflects the sales situation of the sample product at different times before the historical moment.

[0175] Based on the sample live data and the historical sample sales data, an initial prediction model is used to predict the sales trend of the sample product within a sample time period, where the sample time period is a period of time after the historical moment.

[0176] Based on the difference between the predicted sales trend and the actual sales trend, the model parameters of the initial prediction model are adjusted to obtain the sales trend prediction model.

[0177] In one possible implementation, the training unit is used for:

[0178] Based on the difference between the predicted sales trend and the actual sales trend, the initial prediction model is trained to obtain an intermediate prediction model;

[0179] The performance of the intermediate prediction model is evaluated using the objective evaluation function to obtain the evaluation results;

[0180] The model parameters of the intermediate prediction model are optimized based on the evaluation results to obtain the sales trend prediction model.

[0181] In one possible implementation, the apparatus further includes a generation unit:

[0182] The acquisition unit is also used to acquire the inventory data at the current moment;

[0183] The generation unit is used to generate an inventory scheduling strategy based on the sales trend of the target product within a target time period and the inventory data.

[0184] In one possible implementation, the generating unit is used for:

[0185] Based on the sales trend of the target product within the target time period and the inventory data, the inventory identification result is determined;

[0186] If the inventory identification result is the target result, the inventory scheduling strategy is generated by combining the factors affecting inventory scheduling.

[0187] In one possible implementation, the device further includes an early warning unit:

[0188] The early warning unit is used to send an early warning notification if it determines that the sales trend of the target product and the inventory data meet the early warning rules within the target time period.

[0189] As can be seen from the above technical solution, for live-streaming e-commerce scenarios, this application provides a method for real-time prediction of product sales trends over a future period during live-streaming. For any moment in the live-streaming process, such as the current moment, during the live-streaming sales process through the target live-streaming room, real-time live-streaming data for the target product in the target live-streaming room at the current moment is obtained. This real-time live-streaming data reflects the sales situation of the target product under the current interaction data. The current interaction data is the interaction data corresponding to the current moment; therefore, the real-time live-streaming data is the latest status information, reflecting the dynamics and short-term fluctuations of user and market demand. Additionally, historical sales data for the target product can also be obtained. This historical sales data reflects the sales situation of the target product under historical interaction data. The historical interaction data is the interaction data corresponding to different moments before the current moment. Historical sales data is the basis for sales prediction, reflecting user and market demand under general circumstances, i.e., reflecting the long-term trends of users and the market. Thus, by combining real-time live-streaming data and historical sales data, a sales trend prediction model can be used to predict the sales trend of the target product within a target time period. The sales trend prediction model is a model trained on a sample dataset for sales prediction, and the target time period is a period after the current moment. This allows for a comprehensive understanding of long-term market and user trends based on a large amount of historical sales data, while also incorporating real-time live data to reflect market changes and user needs in real time. This improves the accuracy and timeliness of sales trend forecasting, ensuring its real-time nature and accuracy.

[0190] This application also provides a computer device capable of executing a sales trend forecasting method. This computer device may be a terminal. Figure 6 This diagram illustrates the structure of a terminal according to an embodiment of this application. Figure 6 In this example, using a smartphone as the terminal:

[0191] refer to Figure 6 The smartphone includes components such as: a radio frequency (RF) circuit 610, a memory 620, an input unit 630, a display unit 640, a sensor 650, an audio circuit 660, a Wi-Fi module 670, a processor 680, and a power supply 690. The input unit 630 may include a touch panel 631 and other input devices 632, the display unit 640 may include a display panel 641, and the audio circuit 660 may include a speaker 661 and a microphone 662. It is understood that... Figure 6The smartphone structure shown does not constitute a limitation on smartphones and may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0192] The memory 620 can be used to store software programs and modules. The processor 680 executes various functions and data processing of the smartphone by running the software programs and modules stored in the memory 620. The memory 620 may mainly include a program storage area and a data storage area. The program storage area may store the operating system, applications required for at least one function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created based on the use of the smartphone (such as audio data, phonebook, etc.). In addition, the memory 620 may include high-speed random access memory, and may also include non-volatile memory, such as at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0193] The processor 680 is the control center of the smartphone, connecting various parts of the smartphone via various interfaces and lines. It performs various functions and processes data by running or executing software programs and / or modules stored in the memory 620, and by accessing data stored in the memory 620. Optionally, the processor 680 may include one or more processing units; preferably, the processor 680 may integrate an application processor and a modem processor, wherein the application processor mainly handles the operating system, user interface, and applications, and the modem processor mainly handles wireless communication. It is understood that the modem processor may not be integrated into the processor 680.

[0194] In this embodiment, the processor 680 in the smartphone can execute the sales trend prediction method provided in the various embodiments of this application.

[0195] The computer device provided in this application embodiment can also be a server. Please refer to [link / reference]. Figure 7 As shown, Figure 7The diagram illustrates the structure of a server 700 provided in this embodiment. The server 700 can vary significantly depending on its configuration or performance. It may include one or more processors, such as a Central Processing Unit (CPU) 722, a memory 732, and one or more storage media 730 (e.g., one or more mass storage devices) for storing application programs 742 or data 744. The memory 732 and storage media 730 can be temporary or persistent storage. The program stored in the storage media 730 may include one or more modules (not shown in the diagram), each module including a series of instruction operations on the server. Furthermore, the CPU 722 may be configured to communicate with the storage media 730 and execute the series of instruction operations stored in the storage media 730 on the server 700.

[0196] Server 700 may also include one or more power supplies 726, one or more wired or wireless network interfaces 750, one or more input / output interfaces 758, and / or one or more operating systems 741, such as Windows Server. TM Mac OS X TM Unix TM Linux TM FreeBSD TM etc.

[0197] In this embodiment, the central processing unit 722 in the server 700 can execute the sales trend prediction method provided in the various embodiments of this application.

[0198] According to one aspect of this application, a computer-readable storage medium is provided for storing a computer program for performing the sales trend forecasting method described in the foregoing embodiments.

[0199] According to one aspect of this application, a computer program product is provided, comprising a computer program stored in a computer-readable storage medium. A processor of a computer device reads the computer program from the computer-readable storage medium and executes the computer program, causing the computer device to perform the methods provided in various optional implementations of the above embodiments.

[0200] The descriptions of the processes or structures corresponding to the above figures each have their own emphasis. For parts of a process or structure that are not described in detail, please refer to the relevant descriptions of other processes or structures.

[0201] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented, for example, in orders other than those illustrated or described herein. Furthermore, the terms “comprising” and “having,” and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0202] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection between apparatuses or units through some interfaces, and may be electrical, mechanical, or other forms.

[0203] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0204] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0205] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a terminal, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing computer programs, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0206] In this application embodiment, the terms "module" or "unit" refer to a computer program or part of a computer program that has a predetermined function and works with other related parts to achieve a predetermined goal, and can be implemented wholly or partially using software, hardware (such as processing circuitry or memory), or a combination thereof. Similarly, a processor (or multiple processors or memory) can be used to implement one or more modules or units. Furthermore, each module or unit can be part of an overall module or unit that includes the functionality of that module or unit.

[0207] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A sales trend forecasting method, characterized in that, The method includes: During the live sales process through the target live room, real-time live data of the target product in the target live room at the current moment is obtained. The real-time live data is used to reflect the sales situation of the target product under the current interaction data. The current interaction data is the interaction data corresponding to the current moment. Obtain historical sales data of the target product. The historical sales data is used to reflect the sales situation of the target product under historical interaction data. The historical interaction data is the interaction data corresponding to different times before the current time. Based on the real-time live data and the historical sales data, a sales trend prediction model is used to predict the sales trend of the target product within a target time period. The sales trend prediction model is a model trained on a sample dataset for sales prediction, and the target time period is a period of time after the current moment.

2. The method according to claim 1, characterized in that, The step of predicting the sales trend of the target product within a target time period based on the real-time live data and the historical sales data using a sales trend prediction model includes: The real-time live data is used to extract features to obtain a first feature, and the historical sales data is used to extract features to obtain a second feature; Arrange the first and second features in chronological order according to their corresponding time points to obtain time series feature data; Based on the time series feature data, the sales trend prediction model is used to predict the sales trend of the target product within the target time period.

3. The method according to claim 2, characterized in that, The step of predicting the sales trend of the target product within a target time period based on the time series feature data and using the sales trend prediction model includes: Determine the stationarity of the time series feature data; If the stationarity is lower than a preset threshold, the time series feature data is differentially processed by the sales trend prediction model to convert the time series feature data into stationary series feature data. The sales trend prediction model is used to predict the stationary sequence feature data to obtain the sales trend of the target product within the target time period.

4. The method according to claim 1, characterized in that, The step of predicting the sales trend of the target product within a target time period based on the real-time live data and the historical sales data using a sales trend prediction model includes: The sales trend prediction model is updated using the real-time live data to obtain the updated sales trend prediction model; Based on the historical sales data, the updated sales trend prediction model is used to predict the sales trend of the target product within the target time period.

5. The method according to claim 4, characterized in that, The step of predicting the sales trend of the target product within a target time period based on the historical sales data and using the updated sales trend prediction model includes: Based on the historical sales data and the real-time live streaming data, the sales trend of the target product within the target time period is predicted by the updated sales trend prediction model.

6. The method according to claim 1, characterized in that, The sales forecasting model includes multiple forecasting sub-models. Based on the real-time live streaming data and the historical sales data, a sales trend forecasting model is used to predict the sales trend of the target product within a target time period, including: Based on the real-time live data and the historical sales data, predictions are made through the multiple prediction sub-models to obtain the prediction results output by each prediction sub-model. Obtain the weight of each prediction sub-model in the plurality of prediction sub-models; Based on the weights of each prediction sub-model and the corresponding prediction results, the sales trend of the target product within the target time period is determined.

7. The method according to claim 1, characterized in that, The method further includes: During the process of live-streaming sales through historical live-streaming rooms, sample live-streaming data for sample products at the historical moments are obtained. Obtain historical sample sales data for the sample product, which reflects the sales situation of the sample product at different times before the historical moment. Based on the sample live data and the historical sample sales data, an initial prediction model is used to predict the sales trend of the sample product within a sample time period, where the sample time period is a period of time after the historical moment. Based on the difference between the predicted sales trend and the actual sales trend, the model parameters of the initial prediction model are adjusted to obtain the sales trend prediction model.

8. The method according to claim 7, characterized in that, The step of adjusting the model parameters of the initial prediction model based on the difference between the predicted sales trend and the actual sales trend to obtain the sales trend prediction model includes: Based on the difference between the predicted sales trend and the actual sales trend, the initial prediction model is trained to obtain an intermediate prediction model; The performance of the intermediate prediction model is evaluated using the objective evaluation function to obtain the evaluation results; The model parameters of the intermediate prediction model are optimized based on the evaluation results to obtain the sales trend prediction model.

9. The method according to any one of claims 1-8, characterized in that, The method further includes: Obtain the inventory data at the current moment; An inventory scheduling strategy is generated based on the sales trend of the target product within the target time period and the inventory data.

10. The method according to claim 9, characterized in that, The step of generating an inventory scheduling strategy based on the sales trend of the target product within a target time period and the inventory data includes: Based on the sales trend of the target product within the target time period and the inventory data, the inventory identification result is determined; If the inventory identification result is the target result, the inventory scheduling strategy is generated by combining the factors affecting inventory scheduling.

11. The method according to claim 9, characterized in that, The method further includes: If the sales trend of the target product within the target time period and the inventory data are determined to meet the early warning conditions, an early warning notification is sent.

12. A sales trend forecasting device, characterized in that, The device includes an acquisition unit and a prediction unit: The acquisition unit is used to acquire real-time live streaming data of the target product in the target live streaming room at the current moment during the live streaming sales process. The real-time live streaming data is used to reflect the sales situation of the target product under the current interaction data. The current interaction data is the interaction data corresponding to the current moment. The acquisition unit is further configured to acquire historical sales data of the target product, the historical sales data being used to reflect the sales performance of the target product under historical interaction data, and the historical interaction data being the interaction data corresponding to different times before the current time. The prediction unit is used to predict the sales trend of the target product within a target time period based on the real-time live data and the historical sales data using a sales trend prediction model. The sales trend prediction model is a model trained on a sample dataset for sales prediction, and the target time period is a period of time after the current moment.

13. A computer device, characterized in that, The computer device includes a processor and memory: The memory is used to store computer programs and to transfer the computer programs to the processor; The processor is configured to execute the method according to any one of claims 1-11 according to instructions in the computer program.

14. A computer-readable storage medium for storing a computer program that, when executed by a processor, causes the processor to perform the method of any one of claims 1-11.

15. A computer program product comprising a computer program that, when executed by a processor, implements the method of any one of claims 1-11.

Images