Accessory sales prediction method and system, electronic equipment and storage medium

Abstract

The invention relates to the field of engineering machinery, in particular to an accessory sales prediction method and system, electronic equipment and a storage medium. The accessory sales volume prediction method comprises the following steps: acquiring a first classification label of a to-be-predicted first accessory and first sales volume related information of the first accessory; the first sales volume related information is input into a target prediction model associated with the first classification label, a prediction result output by the target prediction model is obtained, and the target prediction model is obtained through training of training samples corresponding to accessories belonging to the first classification label; and according to the prediction result, determining a target sales volume parameter of the first accessory in the future time period. According to the accessory sales volume prediction method provided by the invention, the accuracy of accessory sales volume prediction is improved.

Description

Technical Field

[0001] The present application relates to the field of engineering machinery, and specifically to a method, system, electronic device and storage medium for predicting the sales volume of accessories. Background Art

[0002] With the widespread use of various equipment in the construction and road sectors, parts loss prediction and inventory management have become important technical means to ensure stable equipment operation, improve operation and maintenance efficiency, and enhance service quality. Accurately forecasting parts sales can effectively reduce inventory costs, optimize supply chain management, and improve after-sales response speed.

[0003] In the exemplary technology, historical sales data, maintenance records, warranty data and other information are input into models such as statistical modeling or machine learning methods to predict the sales volume of accessories.

[0004] However, this type of model has high requirements on the linearity and stability of the input data, and the sales patterns of different accessories are different, which makes the sales data corresponding to different accessories nonlinear and volatile, resulting in low accuracy in the sales forecast of accessories. Summary of the Invention

[0005] In view of this, the present application provides a method, system, electronic device and storage medium for predicting sales of accessories to solve the technical problem of low accuracy in sales prediction of accessories.

[0006] In a first aspect, the present application provides a method for predicting accessory sales, comprising:

[0007] Obtaining a first classification label of a first accessory to be predicted and first sales volume related information of the first accessory;

[0008] Inputting the first sales-related information into a target prediction model associated with the first classification label to obtain a prediction result output by the target prediction model, where the target prediction model is trained using training samples corresponding to accessories belonging to the first classification label;

[0009] Based on the prediction result, target sales parameters of the first accessory in a future time period are determined.

[0010] In a possible implementation, obtaining the first classification label of the first accessory to be predicted includes:

[0011] Obtain historical sales parameters of the first accessory in a historical time period;

[0012] Determining, based on the historical sales parameters, the sales frequency, usage cycle, and sales fluctuation parameters of the first accessory in the historical time period;

[0013] Determine a first classification label for the first accessory according to the sales frequency, the usage cycle, and the sales volume fluctuation parameter.

[0014] In a possible implementation, before obtaining the first category label of the first accessory to be predicted and information related to sales volume of the first accessory, the method further includes:

[0015] Obtaining second sales-related information for each second accessory, and generating target time series features of the corresponding second accessory based on the second sales-related information;

[0016] processing target time series features corresponding to the second accessories according to the second classification labels to obtain training samples corresponding to each second classification label, wherein each second classification label includes the first classification label;

[0017] According to the training samples corresponding to each second classification label, the preset model corresponding to the second classification label is trained to obtain a prediction model corresponding to each second classification label, and the prediction model is associated with the corresponding second classification label and stored.

[0018] In a possible implementation, generating the target time series features of the corresponding second accessory according to the second sales-related information includes:

[0019] obtaining a plurality of reference parameters from the second sales-related information corresponding to the second accessory, the plurality of reference parameters including at least two of the following: historical sales volume of the second accessory, number of uses of the device on which the second accessory is installed, service and maintenance times of the second accessory, age of the second accessory, climate and environmental parameters of the region where the second accessory is installed, and sales parameters of a replacement accessory for the second accessory;

[0020] Performing data processing on each of the reference parameters to obtain multiple target parameters;

[0021] Performing time alignment on each of the target parameters according to the time dimension to obtain an initial time series feature corresponding to the second accessory;

[0022] Normalizing the initial time series features to obtain target time series features corresponding to the second accessory.

[0023] In a possible implementation, processing the target time series feature corresponding to the second accessory according to the second classification label to which the second accessory belongs includes:

[0024] When the second classification label to which the second accessory belongs is a fast-flow accessory, a window sliding is performed on the target time series feature corresponding to the second accessory to obtain a plurality of continuous local time series features;

[0025] Performing linear mapping on each of the local time series features to obtain a plurality of first feature vectors;

[0026] A plurality of training samples corresponding to the fast flow accessories are constructed according to each of the first feature vectors.

[0027] In a possible implementation, processing the target time series feature corresponding to the second accessory according to the second classification label to which the second accessory belongs includes:

[0028] When the second classification label to which the second accessory belongs is a medium-flow accessory or a slow-flow accessory, generating a second feature vector according to the target time series feature corresponding to the second accessory;

[0029] A training sample corresponding to the second accessory is constructed according to the second feature vector.

[0030] In a possible implementation, determining the target sales parameter of the first accessory in a future time period based on the prediction result includes:

[0031] The target sales parameter of the first accessory in the future time period and the confidence level corresponding to the target sales parameter are obtained from the prediction result, where the confidence level is used to indicate the prediction accuracy of the target sales parameter.

[0032] As a second aspect of the present application, the present application further provides an accessories sales forecasting system, comprising:

[0033] An acquisition module, configured to acquire a first classification label of a first accessory to be predicted and first sales-related information of the first accessory;

[0034] an input module, configured to input the first sales-related information into a target prediction model associated with the first classification label, and obtain a prediction result output by the target prediction model, wherein the target prediction model is trained using training samples corresponding to accessories belonging to the first classification label;

[0035] A determination module is used to determine the target sales parameters of the first accessory in a future time period based on the prediction result.

[0036] As a third aspect of the present application, the present application further provides an electronic device, including a memory and a processor, wherein:

[0037] The memory is connected to the processor and is used to store programs;

[0038] The processor is used to implement the above-mentioned accessory sales prediction method by running the program in the memory.

[0039] As a fourth aspect of the present application, the present application further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the accessory sales prediction method as described above is implemented.

[0040] The present application provides a method, system, electronic device and storage medium for predicting accessory sales, wherein the accessory sales prediction method includes the following steps: obtaining the classification label and sales-related information of the accessory to be predicted, inputting the sales-related information into the prediction model associated with the classification label, and obtaining the sales parameters of the accessory in the future time period. The accessory sales prediction method provided in the present application inputs the sales-related information of the accessory into the prediction model corresponding to the classification label to which the accessory belongs, and the prediction model is obtained by training the training samples of each accessory under the classification label, that is, by learning the exclusive model of the sales law of the accessory under the classification label, the sales of the accessory to be tested under the classification label in the future time period are accurately predicted, thereby improving the accuracy of the accessory sales prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] The above and other purposes, features, and advantages of the present application will become more apparent through a more detailed description of the embodiments of the present application in conjunction with the accompanying drawings. The accompanying drawings are intended to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the drawings, the same reference numerals generally represent the same components or steps.

[0042] Figure 1 Shown is a scenario diagram of the accessories sales forecasting method involved in this application.

[0043] Figure 2 The figure shows one of the flow charts of the accessory sales forecasting method provided in one embodiment of the present application.

[0044] Figure 3 The figure shows the second flow chart of the accessory sales forecasting method provided by another embodiment of the present application.

[0045] Figure 4 The figure shows the third flow chart of the accessory sales forecasting method provided in another embodiment of the present application.

[0046] Figure 5 Shown is a fourth flow chart of the accessory sales forecasting method provided in another embodiment of the present application.

[0047] Figure 6The figure shows a brief flow chart of an accessory sales forecasting method provided in another embodiment of the present application.

[0048] Figure 7 Shown is a structural block diagram of an accessories sales forecasting system provided in one embodiment of the present application.

[0049] Figure 8 Shown is a structural block diagram of an electronic device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0050] In the description of the application, the meaning of "multiple" is at least two, for example two, three, etc., unless otherwise clearly and specifically limited. In the embodiments of the present application, all directional indications (such as up, down, left, right, front, back, top, bottom ...) are only used to explain the relative position relationship, motion situation, etc. between each component under a certain specific posture (as shown in the drawings). If this specific posture changes, this directional indication also changes accordingly. In addition, the terms "comprise" and "have" and any deformation thereof are intended to cover non-exclusive inclusion. For example, the process, method, system, product or equipment comprising a series of steps or units is not limited to the steps or units listed, but optionally also includes the steps or units not listed, or optionally also includes other steps or units inherent to these processes, methods, products or equipment.

[0051] In addition, references to "embodiments" herein mean that a particular feature, structure, or characteristic described in connection with the embodiments may be included in at least one embodiment of the present application. The appearance of such phrases in various places in the specification does not necessarily refer to the same embodiment, nor does it necessarily refer to independent or alternative embodiments that are mutually exclusive of other embodiments. It is understood, both explicitly and implicitly, by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0052] Application Overview

[0053] With regard to the technical problem of low accuracy in sales forecasting of accessories in the exemplary technology, the present application further analyzed and found that:

[0054] Different types of accessories have different sales patterns in the market. For example, one type of accessories has large sales and significant fluctuations, another type of accessories has obvious sales patterns and moderate fluctuations, and another type of accessories has sparse sales. The model learns the sales patterns of these accessories at the same time, causing the model to overfit. As a result, the model cannot accurately capture the sales patterns of these types of accessories, resulting in poor accuracy in predicting the sales of these types of accessories.

[0055] In view of this, the present application provides a method, system, electronic device and storage medium for predicting the sales volume of accessories. The accessory sales prediction method obtains the classification label and sales-related information of the accessory to be predicted, inputs the sales-related information into the prediction model associated with the classification label, and obtains the sales parameters of the accessory in the future time period. The accessory sales prediction method provided in the present application inputs the sales-related information of the accessory into the prediction model corresponding to the classification label to which the accessory belongs, and the prediction model is obtained by training the training samples of each accessory under the classification label, that is, by learning the exclusive model of the sales law of the accessory under the classification label, the sales volume of the accessory to be tested under the classification label in the future time period is accurately predicted, thereby improving the accuracy of the sales prediction of the accessory.

[0056] Reference Figure 1 , Figure 1 This is a scenario diagram of the accessory sales forecasting method for this application. Figure 1 As shown, a plurality of prediction models are provided in the accessory sales prediction system, such as prediction model 110, prediction model 120, and prediction model 130. Each prediction model is associated with a classification label, and the classification label refers to the type of accessory. The accessory sales prediction system obtains the classification label of the accessory to be predicted and sales-related information. For example, an external device or a user inputs the classification label and sales-related information into the accessory sales prediction system. The accessory sales prediction system determines the prediction model corresponding to the classification label. For example, the model corresponding to the classification label is prediction model 130. The accessory sales prediction system inputs the sales-related information into the prediction model 130, and the target sales parameters of the accessory in the non-opening time period output by the prediction model 130 are obtained. The accessory sales prediction system outputs the target sales parameters.

[0057] The following will be combined Figure 1 The accompanying drawings in the embodiments of this application clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the described embodiments are only part of the embodiments of this application, not all of them. All other embodiments obtained by ordinary technicians in this field based on the embodiments of this application without making any creative efforts shall fall within the scope of protection of this application.

[0058] Exemplary Methods

[0059] Figure 2 The figure shows one of the flow charts of a method for predicting the sales volume of accessories provided by this application. Figure 2 As shown, this accessory sales forecasting method may specifically include the following steps:

[0060] Step S210: Obtain a first classification label of a first accessory to be predicted and first sales volume related information of the first accessory.

[0061] In this embodiment, the execution entity is a parts sales forecasting system. For ease of description, the term "system" will be used to refer to the parts sales forecasting system. The system can be composed of a server, computer, or terminal device equipped with parts sales forecasting capabilities. Parts can be components of any device, for example, components in construction equipment.

[0062] The system is equipped with multiple prediction models, each used to predict the sales volume of different types of accessories. Each accessory has a corresponding classification label that characterizes the type of accessory. Exemplarily, the classification labels include fast-flow accessories, medium-flow accessories, and slow-flow accessories. That is, accessories are categorized by their turnover rate. For example, a fast-flow accessory is one with a turnover rate greater than a first preset rate; a medium-flow accessory is one with a turnover rate greater than a second preset rate and less than or equal to the first preset rate; and a slow-flow accessory is one with a turnover rate less than or equal to the second preset rate, where the second preset rate is less than the first preset rate.

[0063] The system obtains a classification label of the accessory to be predicted, the accessory to be predicted is defined as a first accessory, and the classification label of the first accessory is defined as a first classification label. The first classification label can be input to the system by a user.

[0064] After obtaining the first classification label of the first accessory, the system obtains the sales-related information of the first accessory, which is defined as the first sales-related information. The first sales-related information is multi-source data, including the historical sales data of the first accessory, the usage of the equipment where the first accessory is located, the service and maintenance data of the first accessory, the service life of the first accessory, the climate and environmental data of the area where the used equipment where the first accessory is located, and at least two of the sales of replacement accessories for the first accessory. The usage of the equipment where the first accessory is located is the equipment holdings, and the equipment holdings refer to the total number of equipment where the first accessory is located that has been put into use and is still in use in a certain area or market. Replacement accessories for the first accessory refer to other types of accessories that are functionally equivalent to the first accessory and can be used interchangeably. For example, if the first accessory is sealing ring 201, the replacement accessory can be sealing ring X01.

[0065] In step S220 , the first sales-related information is input into a target prediction model associated with the first classification label to obtain a prediction result output by the target prediction model, where the target prediction model is trained by various training samples corresponding to accessories belonging to the first classification label.

[0066] The system obtains the first sales-related information and inputs the first sales-related information into the target prediction model associated with the first classification label. The target prediction model predicts the sales of the first accessory in the future time period based on the first sales-related information, that is, the target prediction model outputs the prediction result.

[0067] Step S230: Determine the target sales volume parameters of the first accessory in the future time period based on the prediction result.

[0068] After obtaining the forecast results, the system determines the target sales parameters for the first accessory in a future time period based on the forecast results. The future time period can refer to a specific time period, for example, the next month following the current month. In one example, the forecast results include the forecast sales volume for the first accessory, and the forecast sales volume in the forecast results is used as the target sales parameters.

[0069] In another example, the forecast result includes predicted sales and a confidence level. The confidence level can be added to the predicted sales by the target forecast model through error analysis or uncertainty assessment methods. The system obtains the predicted sales of the first accessory in the future time period from the forecast result as the target sales parameter, and also obtains the confidence level of the target sales parameter from the forecast result. The confidence level indicates the accuracy of the target sales parameter prediction. The higher the confidence level, the more accurate the prediction of the target sales parameter.

[0070] In this embodiment, the classification label and sales-related information of the accessory to be predicted are obtained, and the sales-related information is input into the prediction model associated with the classification label to obtain the sales parameters of the accessory in the future time period. The accessory sales prediction method provided in this embodiment inputs the sales-related information of the accessory into the prediction model corresponding to the classification label to which the accessory belongs. The prediction model is trained using training samples of each accessory under the classification label. In other words, by learning the sales patterns of accessories under the classification label, a dedicated model is used to accurately predict the sales of the accessory under the classification label in the future time period, thereby improving the accuracy of accessory sales prediction.

[0071] Reference Figure 3 , Figure 3 This is a flow chart of a method for predicting the sales volume of accessories provided in this application, based on the following Figure 2 In the illustrated embodiment, step S210 includes:

[0072] Step S310: Obtain historical sales parameters of the first accessory in a historical time period.

[0073] Step S320: determining the sales frequency, usage cycle, and sales fluctuation parameters of the first accessory in a historical time period based on the historical sales parameters.

[0074] Step S330: Determine a first classification label for the first accessory based on the sales frequency, usage cycle, and sales volume fluctuation parameters.

[0075] In this embodiment, the system obtains historical sales parameters of the first accessory in a historical time period, and determines the first category label to which the first accessory belongs based on the historical sales parameters.

[0076] For example, the system determines the sales frequency, usage cycle, and sales fluctuation parameter of the first accessory during a historical time period based on historical sales parameters. The sales frequency refers to the sales volume of the first accessory during the historical time period or the ratio of sales volume to the duration of the historical time period. The usage cycle refers to the interval between the start time and replacement time of the first accessory. The sales fluctuation parameter refers to the sales difference between two time periods, or the fluctuation exhibited by multiple sales differences. For example, if the multiple sales differences are uneven, the sales fluctuation parameter indicates large sales fluctuations.

[0077] After obtaining the sales frequency, usage cycle, and sales volume fluctuation parameters, the system determines a first classification label for the first accessory based on the sales frequency, usage cycle, and sales volume fluctuation parameters.

[0078] For example, if the sales frequency of the first accessory is greater than a first preset threshold, the sales fluctuation pattern of the first accessory indicated by the sales fluctuation parameter is obvious, and the usage cycle is less than a preset cycle, then the first classification label is a fast-flow accessory;

[0079] If the sales frequency of the first accessory is less than or equal to a first preset threshold and greater than a second preset threshold, the sales fluctuation parameter indicates that the sales fluctuation is complex and presents a certain sales fluctuation, and the usage period is greater than a preset period, then the first classification label is a mid-stream accessory;

[0080] If the sales frequency of the first accessory is less than or equal to or greater than the second preset threshold, the sales fluctuation parameter indicates that the sales data of the first accessory is limited and the usage cycle is difficult to count, that is, the relevant data of the first accessory is relatively small, then the first classification label of the first accessory is a slow-flow accessory.

[0081] In this embodiment, the system accurately determines the first category label to which the first accessory belongs based on the historical sales parameters of the first accessory.

[0082] Reference Figure 4 , Figure 4 This is a flowchart of a method for predicting the sales volume of accessories provided in this application, based on the following Figure 2 or Figure 3 In the embodiment shown, before step S210, the following steps are further included:

[0083] Step S410 , obtaining second sales-related information of each second accessory, and generating target time series features of the corresponding second accessory according to the second sales-related information.

[0084] In this embodiment, the system obtains second sales-related information for each second accessory, and the second sales-related information includes the historical sales quantity of the second accessory, the number of equipment uses corresponding to the equipment where the second accessory is located, the number of service and maintenance times of the second accessory, the service life of the second accessory, the climate and environmental parameters of the area where the second accessory is located, and at least two of the sales parameters of the replacement accessories of the second accessory.

[0085] The second sales-related information includes time, and the second sales-related information is feature integrated according to the time dimension to obtain the target time series features of the second accessory.

[0086] Step S420 : processing the target time series features corresponding to the second accessories according to the second classification labels to which the second accessories belong, and obtaining training samples corresponding to each second classification label, wherein each second classification label includes the first classification label.

[0087] Each second accessory has a corresponding second classification label. Different second classification labels process the target time series features in different ways. After processing the target time series, the feature vector corresponding to the second accessory can be obtained. The feature vector of the second accessory and the second classification label construct the training sample corresponding to the second accessory.

[0088] In one example, different second classification labels are set with corresponding processing strategies. The system obtains the second classification label to which the second accessory belongs, and performs a strategy on the target time strategy corresponding to the second accessory based on the processing strategy associated with the second classification label to obtain a feature vector corresponding to the second accessory.

[0089] In another example, when the second classification label to which the second accessory belongs is a fast-flow accessory, the target time series feature corresponding to the second accessory is window-sliding to obtain a plurality of continuous local time series features. For example, a fixed window length (such as 12 months) is used to perform sliding segmentation on the target time series feature of the second accessory to obtain a plurality of continuous local time series features. After obtaining each local time series feature, each local time series feature is linearly mapped to obtain a plurality of first feature vectors, thereby constructing a plurality of training samples of fast-flow accessories based on each first feature vector. Exemplarily, the local time series feature contains features of multiple dimensions, and the local time series feature is embedded into a unified dimensional space through linear mapping, and the position encoding vector is superimposed and the time sequence information is retained to obtain the first feature vector corresponding to the local time series feature.

[0090] When the second classification label to which the second accessory belongs is a medium-flow accessory or a slow-flow accessory, a second feature vector is generated based on the target time series feature corresponding to the second accessory, and a training sample corresponding to the second accessory is constructed based on the second feature vector.

[0091] Step S430: train the preset model corresponding to each second classification label according to the training samples corresponding to each second classification label, obtain the prediction model corresponding to each second classification label, and associate the prediction model with the corresponding second classification label for storage.

[0092] Based on the training samples corresponding to the second classification label of the breeding, the system trains the preset model corresponding to the second classification label to obtain the prediction model corresponding to each second classification label, and then associates the prediction model with the corresponding second classification label for storage.

[0093] For example, if the second classification label indicates that the second accessory is a fast-flow accessory, a time series block transformer can be used as the preset model for the fast-flow accessory. The system inputs the first feature vector into the encoder of the preset model, and uses the multi-head attention mechanism and the fully connected layer in the preset model to extract key timing dependencies and feature interactions, and finally outputs the sales forecast value for the future time period. The preset model adopts a supervised learning method, and the prediction target is the sales value of one or more time periods. The loss function can use the mean square error, and the network weight parameters of the preset model are continuously optimized through back propagation.

[0094] If the second classification label indicates that the second accessory is a mid-stream accessory, a long short-term memory network with a recurrent neural network architecture is used to construct a preset model corresponding to the mid-stream accessory, and the second feature vector corresponding to the mid-stream accessory is input into the preset model. The preset model corresponding to the mid-stream accessory can be trained to obtain a prediction model associated with the mid-stream accessory.

[0095] If the second classification label indicates that the second accessory is a slow-flow accessory, since the sales samples of slow-flow accessories are sparse and the historical data is limited, it is difficult to support deep model training. Therefore, the gradient boosting regression model based on the tree model is used as the preset model corresponding to the slow-flow accessory. The system uses the second feature vector of the slow-flow accessory as a training sample and inputs it into the preset model for training. During the training process, the incremental learning of each round of the model is completed based on the residual iteration mechanism in the preset model to improve the fitting accuracy of the prediction model for slow-flow accessories. Furthermore, the second feature vector is a feature vector constructed by time dimension features. The sliding window average sales volume, equipment replacement cycle and other features can be added to the second feature vector of the slow-flow accessory to obtain the target feature vector, and the preset model is trained based on the target feature vector. The output of the trained prediction model is the estimated sales volume in the future period, and the prediction result is highly stable and computationally efficient. The system can deploy the prediction model corresponding to the slow-flow accessory on resource-constrained devices to perform prediction tasks.

[0096] In this embodiment, as training samples for the training model, multiple types of data such as accessory sales, equipment ownership, service and maintenance records, years of use, climate factors, and substitution relationships are integrated to construct a relatively complete feature system, providing a reliable data basis for the training of the prediction model. In addition, according to the different circulation speeds of accessories, the second accessories are divided into second classification labels such as fast-flow accessories, medium-flow accessories, and slow-flow accessories. Based on different second classification labels, different models are used for training to better adapt to the actual demand characteristics of various types of accessories, that is, the model is trained by type to improve the applicability of sales prediction for different accessories. In addition, the introduction of time training modeling enhances the trend recognition of the prediction model, and the use of structures such as Transformer and recurrent neural networks to model sales sequences can more effectively capture sales change trends and improve prediction accuracy.

[0097] Reference Figure 5 , Figure 5 This is a flowchart of a method for predicting the sales volume of accessories provided in this application, based on the following Figure 4 In the illustrated embodiment, step S410 includes:

[0098] Step S510, obtaining multiple reference parameters from the second sales-related information corresponding to the second accessory, the multiple reference parameters including the historical sales quantity of the second accessory, the number of equipment uses corresponding to the equipment where the second accessory is located, the number of service and maintenance times of the second accessory, the service life of the second accessory, the climate and environmental parameters of the area where the second accessory is located, and at least two of the sales parameters of the replacement accessories of the second accessory.

[0099] In this embodiment, the device obtains multiple reference parameters from the second sales-related information corresponding to the second accessory, and the multiple reference parameters include the historical sales quantity of the second accessory, the number of equipment uses corresponding to the equipment where the second accessory is located, the number of service and maintenance times of the second accessory, the service life of the second accessory, the climate and environmental parameters of the area where the second accessory is located, and at least two of the sales parameters of the replacement accessories of the second accessory.

[0100] Step S520: performing data processing on each reference parameter to obtain multiple target parameters.

[0101] After obtaining the reference parameters, data processing is performed on the reference parameters.

[0102] For example, accessories are classified by model, and their actual sales quantities in different regions are summarized on a monthly basis; the number of corresponding equipment in each region and service and maintenance records, including the number of repairs and maintenance, are simultaneously counted; the design service life and recommended replacement cycle of each type of accessories are extracted; the sales volume of alternative accessories in the same time dimension is summarized; and climate indicators such as the monthly average temperature, monthly average humidity, monthly average precipitation, monthly average sunshine duration and monthly average air pressure in the corresponding region are obtained from the meteorological data platform.

[0103] The device performs an integrity check on the aggregated data set. For missing feature values, it uses a regression model based on the gradient boosting tree algorithm to predict and fill in the missing items. It uses other relevant features to train the regressor to estimate the missing items to ensure the integrity and accuracy of the data.

[0104] The device uses the box plot method to perform statistical analysis on each type of numerical feature, calculates the upper and lower quartiles of the values ​​and the abnormal threshold range, removes outliers that exceed the threshold, or replaces them with the historical median of this type of feature to avoid interference of extreme values ​​on model training.

[0105] Step S530 : Time-aligning each target parameter according to the time dimension to obtain an initial time series feature corresponding to the second accessory.

[0106] After the device processes the reference parameters, it can obtain the target parameters. The device aligns the target parameters according to the time dimension to obtain the initial time series characteristics corresponding to the second accessory.

[0107] For example, the processed multidimensional data is aligned and organized by time (e.g., monthly granularity) to generate a time series dataset with a unified structure. Each time series sample contains the historical sales volume, inventory, age vector, climate indicator, and replacement part sales volume of the current accessory. This time series sample is the initial time series feature.

[0108] Step S540 : normalize the initial time series features to obtain target time series features corresponding to the second accessory.

[0109] After obtaining the initial time series features, all numerical features in the initial time series are normalized and scaled to the [0, 1] interval using the minimum-maximum normalization method to ensure the scale consistency of each feature during the model training process and improve the model convergence efficiency.

[0110] Based on the above embodiment, refer to Figure 6 , briefly describe this application.

[0111] In this application, multi-source data of accessories is obtained, and the multi-source data is uniformly collected and preliminarily processed. For example, monthly sales records and service maintenance records of accessories are obtained from the equipment management platform, the number of equipment in each region and model is counted, and the design service life and replacement cycle of accessories are extracted; at the same time, climate indicators such as the monthly average temperature, humidity, precipitation, sunshine duration and air pressure in the target area are obtained through a third-party meteorological interface; and an accessories substitution relationship table is established to clarify the substitution ratio between interchangeable accessories. After missing values ​​are filled, outliers are removed, normalized and encoded, all raw data form a multidimensional time series input aligned to the monthly granularity.

[0112] In the feature construction stage, the "unit equipment demand intensity" indicator is first calculated, that is, the ratio of the sales volume of accessories in each month to the corresponding equipment holdings; then the periodicity and volatility characteristics of the time series are extracted through the sliding window method; the accessory service life factor is introduced to express its replacement trend within its life cycle; the climate data is aligned with the time series to generate external environmental characteristics such as monthly average temperature, humidity, precipitation, sunshine, and air pressure; the accessory service and maintenance data is integrated to extract key maintenance behaviors (such as replacement frequency, first replacement time, and average maintenance cycle) as important indicators reflecting the actual use of accessories; the accessory substitution relationship is converted into a graph structure, and a graph neural network is used to calculate the functional similarity embedding vector of each accessory. Finally, all dimensional features are used as model input.

[0113] At the modeling level, first, based on the average monthly sales of each accessory over the past x months, they are divided into three categories: fast-moving accessories, medium-moving accessories, and slow-moving accessories:

[0114] Fast-flow software: Suitable for large amounts of data with significant fluctuations. It uses the Patch Time Series Transformer (PatchTST) model for training and prediction.

[0115] Mid-stream parts: Sales have certain regularity but moderate fluctuations. Long short-term memory (LSTM) network models are used for training and prediction.

[0116] Slow-flow items: Sales are sparse and the sample size is limited. The gradient boosting tree model XGBoost is used for training and prediction.

[0117] Each model takes the multidimensional time series features of the corresponding category as input, uses the mean square error (MSE) loss function for supervised training, and outputs the spare parts demand and its confidence interval in one or more future time periods in the forecast stage.

[0118] This application can modularly deploy the above method on servers or edge devices to form an accessory sales forecasting system to support on-demand and batch forecasting tasks. The accessory sales forecasting method provided in this application takes into account prediction accuracy, operational efficiency, and deployment flexibility, and is suitable for various scenarios such as equipment accessory management, production planning, and after-sales inventory scheduling.

[0119] Corresponding to the above-mentioned accessory sales forecasting method, this application also provides an accessory sales forecasting system. Figure 7 The accessories sales forecasting system is described in detail.

[0120] like Figure 7 As shown, the accessory sales forecasting system 700 includes:

[0121] An acquisition module 710 is configured to acquire a first classification label of a first accessory to be predicted and first sales volume related information of the first accessory;

[0122] An input module 720 is configured to input the first sales-related information into a target prediction model associated with the first classification label to obtain a prediction result output by the target prediction model, where the target prediction model is trained using training samples corresponding to accessories belonging to the first classification label.

[0123] The determination module 730 is configured to determine a target sales parameter of the first accessory in a future time period based on the prediction result.

[0124] In one embodiment, the accessory sales forecasting system 700 is further configured to:

[0125] Obtain historical sales parameters of the first accessory in a historical time period;

[0126] Determine the sales frequency, usage cycle, and sales fluctuation parameters of the first accessory in a historical time period based on the historical sales parameters;

[0127] A first classification label for the first accessory is determined according to sales frequency, usage cycle, and sales volume fluctuation parameters.

[0128] In one embodiment, the accessory sales forecasting system 700 is further configured to:

[0129] Obtaining second sales-related information for each second accessory, and generating target time series features of the corresponding second accessory based on the second sales-related information;

[0130] According to the second classification labels of the second accessories, target time series features corresponding to the second accessories are processed to obtain training samples corresponding to each second classification label, where each second classification label includes the first classification label;

[0131] According to the training samples corresponding to each second classification label, the preset model corresponding to the second classification label is trained to obtain a prediction model corresponding to each second classification label, and the prediction model is associated with the corresponding second classification label and stored.

[0132] In one embodiment, the accessory sales forecasting system 700 is further configured to:

[0133] obtaining a plurality of reference parameters from the second sales-related information corresponding to the second accessory, the plurality of reference parameters including at least two of the following: historical sales volume of the second accessory, usage volume of the device on which the second accessory is installed, service and maintenance times of the second accessory, service life of the second accessory, climate and environmental parameters of the region where the second accessory is installed, and sales parameters of a replacement accessory for the second accessory;

[0134] Perform data processing on each reference parameter to obtain multiple target parameters;

[0135] According to the time dimension, each target parameter is time-aligned to obtain the initial time series features corresponding to the second accessory;

[0136] The initial time series features are normalized to obtain the target time series features corresponding to the second accessory.

[0137] In one embodiment, the accessory sales forecasting system 700 is further configured to:

[0138] When the second classification label to which the second accessory belongs is a fast-flow accessory, a window sliding is performed on the target time series features corresponding to the second accessory to obtain multiple continuous local time series features;

[0139] Perform linear mapping on each local time series feature to obtain multiple first eigenvectors;

[0140] Based on each first eigenvector, a plurality of training samples corresponding to the fast flow accessory are constructed.

[0141] In one embodiment, the accessory sales forecasting system 700 is further configured to:

[0142] When the second classification label to which the second accessory belongs is a medium-flow accessory or a slow-flow accessory, generating a second feature vector according to the target time series feature corresponding to the second accessory;

[0143] A training sample corresponding to the second accessory is constructed according to the second eigenvector.

[0144] In one embodiment, the accessory sales forecasting system 700 is further configured to:

[0145] The target sales parameter of the first accessory in the future time period and the confidence level corresponding to the target sales parameter are obtained from the prediction result. The confidence level is used to indicate the prediction accuracy of the target sales parameter.

[0146] Below, reference Figure 8 To describe the electronic device according to the embodiment of the present application.

[0147] Figure 8 The figure shows a structural block diagram of an electronic device according to an embodiment of the present application.

[0148] like Figure 8 As shown, electronic device 800 includes one or more processors 810 and memory 820 .

[0149] The processor 810 may be a central processing unit (CPU) or other forms of processing units having data processing capabilities and / or instruction execution capabilities, and may control other components in the electronic device 800 to perform desired functions.

[0150] The memory 820 may include one or more computer program products, which may include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. Volatile memory may include, for example, random access memory (RAM) and / or cache memory. Non-volatile memory may include, for example, read-only memory (ROM), a hard disk, flash memory, etc. One or more computer program instructions may be stored on the computer-readable storage medium, and the processor 11 may execute the program instructions to implement the multi-variable expression calculation method of each embodiment of the present application described above and / or other desired functions.

[0151] In one example, the electronic device 800 may further include an input device 830 and an output device 840 , and these components are interconnected via a bus system and / or other forms of connection mechanisms (not shown).

[0152] When the electronic device is a stand-alone device, the input device 830 may be a communication network connector, configured to receive collected input signals from the first device and the second device.

[0153] In addition, the input device 830 may also include, for example, a keyboard, a mouse, and the like.

[0154] The output device 840 can output various information to the outside, including determined distance information, direction information, etc. The output device 840 can include, for example, a display, a speaker, a printer, a communication network and its connected remote output device, etc.

[0155] Of course, to simplify, Figure 8Only some of the components related to the present application in the electronic device 800 are shown, and components such as buses, input / output interfaces, etc. are omitted. In addition, the electronic device 800 may further include any other appropriate components according to specific application scenarios.

[0156] The present application also provides a computer-readable storage medium, wherein the storage medium stores a computer program, and the computer program is configured to perform the following steps:

[0157] Obtaining a first classification label of a first accessory to be predicted and first sales volume related information of the first accessory;

[0158] Inputting the first sales-related information into a target prediction model associated with the first classification label to obtain a prediction result output by the target prediction model, where the target prediction model is trained using training samples corresponding to accessories belonging to the first classification label;

[0159] Based on the forecast results, the target sales parameters of the first accessory in the future time period are determined.

[0160] In addition to the above-mentioned methods and devices, an embodiment of the present application may also be a computer program product, which includes computer program information, which, when executed by a processor, enables the processor to execute the steps of the accessory sales prediction method according to various embodiments of the present application described in this specification.

[0161] The computer program product may be written in any combination of one or more programming languages ​​to implement the program code for performing the operations of the embodiments of the present application, including object-oriented programming languages ​​such as Java, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computing device, partially on the user's computing device, as a standalone software package, partially on the user's computing device and partially on a remote computing device, or entirely on a remote computing device or server.

[0162] In addition, an embodiment of the present application may also be a computer-readable storage medium on which computer program information is stored. When the computer program information is executed by a processor, the processor executes the steps in the accessory sales prediction method according to various embodiments of the present application described in this specification.

[0163] Computer readable storage media can adopt any combination of one or more readable media. The readable medium can be a readable signal medium or a readable storage medium. The readable storage medium can, for example, include but is not limited to a system, device or component of electricity, magnetism, light, electromagnetic, infrared, or semiconductor, or any combination thereof. More specific examples (non-exhaustive list) of readable storage media include: an electrical connection with one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof.

[0164] The basic principles of the present application have been described above in conjunction with specific embodiments. However, it should be noted that the advantages, strengths, and effects mentioned in this application are merely illustrative and not restrictive, and it should not be assumed that these advantages, strengths, and effects are required of each embodiment of this application. In addition, the specific details disclosed above are merely illustrative and facilitating understanding, and are not restrictive. The above details do not limit this application to necessarily being implemented using the above specific details.

[0165] The block diagrams of the devices, devices, equipment, and systems involved in this application are merely illustrative examples and are not intended to require or imply that they must be connected, arranged, or configured in the manner shown in the block diagrams. As will be appreciated by those skilled in the art, these devices, devices, equipment, and systems can be connected, arranged, or configured in any manner. Words such as "include," "comprise," "have," and the like are open-ended words, meaning "including but not limited to," and can be used interchangeably therewith. The words "or" and "and" used herein refer to the words "and / or" and can be used interchangeably therewith, unless the context clearly indicates otherwise. The word "such as" used herein refers to the phrase "such as but not limited to," and can be used interchangeably therewith.

[0166] It should also be noted that in the apparatus, device, and method of the present application, each component or each step can be decomposed and / or recombined, and such decomposition and / or recombination should be regarded as equivalent solutions of the present application.

Claims

1. A method for predicting sales volume of accessories, characterized in that: include: Obtaining a first classification label of a first accessory to be predicted and first sales volume related information of the first accessory; Inputting the first sales-related information into a target prediction model associated with the first classification label to obtain a prediction result output by the target prediction model, where the target prediction model is trained using training samples corresponding to accessories belonging to the first classification label; Based on the prediction result, target sales parameters of the first accessory in a future time period are determined.

2. The accessory sales forecasting method according to claim 1, characterized in that: The obtaining of the first classification label of the first accessory to be predicted includes: Obtain historical sales parameters of the first accessory in a historical time period; Determining, based on the historical sales parameters, the sales frequency, usage cycle, and sales fluctuation parameters of the first accessory in the historical time period; Determine a first classification label for the first accessory according to the sales frequency, the usage cycle, and the sales volume fluctuation parameter.

3. The accessory sales forecasting method according to claim 1, characterized in that: Before obtaining the first classification label of the first accessory to be predicted and the sales volume related information of the first accessory, the method further includes: Obtaining second sales-related information for each second accessory, and generating target time series features of the corresponding second accessory based on the second sales-related information; processing target time series features corresponding to the second accessories according to the second classification labels to obtain training samples corresponding to each second classification label, wherein each second classification label includes the first classification label; According to the training samples corresponding to each second classification label, the preset model corresponding to the second classification label is trained to obtain a prediction model corresponding to each second classification label, and the prediction model is associated with the corresponding second classification label and stored.

4. The accessory sales forecasting method according to claim 3, characterized in that: Generating the target time series features of the corresponding second accessory according to the second sales-related information includes: obtaining a plurality of reference parameters from the second sales-related information corresponding to the second accessory, the plurality of reference parameters including at least two of the following: historical sales volume of the second accessory, number of uses of the device on which the second accessory is installed, service and maintenance times of the second accessory, age of the second accessory, climate and environmental parameters of the region where the second accessory is installed, and sales parameters of a replacement accessory for the second accessory; Performing data processing on each of the reference parameters to obtain multiple target parameters; Performing time alignment on each of the target parameters according to the time dimension to obtain an initial time series feature corresponding to the second accessory; Normalizing the initial time series features to obtain target time series features corresponding to the second accessory.

5. The accessory sales forecasting method according to claim 3, characterized in that: The processing of the target time series feature corresponding to the second accessory according to the second classification label to which the second accessory belongs includes: When the second classification label to which the second accessory belongs is a fast-flow accessory, a window sliding is performed on the target time series feature corresponding to the second accessory to obtain a plurality of continuous local time series features; Performing linear mapping on each of the local time series features to obtain a plurality of first feature vectors; A plurality of training samples corresponding to the fast flow accessories are constructed according to each of the first feature vectors.

6. The accessory sales forecasting method according to claim 3, characterized in that: The processing of the target time series feature corresponding to the second accessory according to the second classification label to which the second accessory belongs includes: When the second classification label to which the second accessory belongs is a medium-flow accessory or a slow-flow accessory, generating a second feature vector according to the target time series feature corresponding to the second accessory; A training sample corresponding to the second accessory is constructed according to the second feature vector.

7. The accessory sales forecasting method according to any one of claims 1 to 6, characterized in that: Determining target sales parameters of the first accessory in a future time period based on the prediction result includes: The target sales parameter of the first accessory in the future time period and the confidence level corresponding to the target sales parameter are obtained from the prediction result, where the confidence level is used to indicate the prediction accuracy of the target sales parameter.

8. A parts sales forecasting system, characterized in that: include: An acquisition module, configured to acquire a first classification label of a first accessory to be predicted and first sales-related information of the first accessory; an input module, configured to input the first sales-related information into a target prediction model associated with the first classification label, and obtain a prediction result output by the target prediction model, wherein the target prediction model is trained using training samples corresponding to accessories belonging to the first classification label; A determination module is used to determine the target sales parameters of the first accessory in a future time period based on the prediction result.

9. An electronic device, characterized in that: comprising a memory and a processor, wherein, The memory is connected to the processor and is used to store programs; The processor is configured to implement the accessory sales prediction method according to any one of claims 1 to 7 by running the program in the memory.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, and when the computer program is executed by the processor, the method for predicting accessory sales volume according to any one of claims 1 to 7 is implemented.

Images