A sales target filling auxiliary method, device, equipment and readable storage medium

Abstract

The present application provides a sales target filling auxiliary method, device, equipment and readable storage medium, which detects and obtains the filled-in sales target name information on the sales target filling page; extracts keywords in the sales target name information; searches for the target historical sales data and target filling requirements in the pre-built sales data knowledge base based on the keywords; generates filling suggestions and reference cases for the sales target name information through a pre-trained large model based on the historical sales data; presents the filling requirements, filling suggestions and reference cases on the target filling page to assist in filling in the sales target elements. The present invention can give filling requirements, filling suggestions and reference cases based on the input sales target name information, and help users understand the filling rules, purpose and reasonable formulation scope of sales target elements from multi-faceted information, which significantly improves the work efficiency and accuracy in the target filling process.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a sales target filling auxiliary method, device, equipment and readable storage medium. Background Art

[0002] In today's digital age, companies use digital software to manage their employees, often requiring employees or users to fill out a large amount of information, and many of the contents are highly professional and difficult for ordinary users to understand. When filling out complex forms, they often do not understand the meaning or function. For example, in a sales target management system, users are required to set three weights, under which many sub-weights must be set, and the sub-weights contain many sales targets, such as target values, disassembly plans, etc. It is difficult for users to understand the meaning and complex logical relationships.

[0003] In the prior art, when filling in complex form contents, only brief information or obscure prompts are provided, and these contents are neither intuitive nor comprehensive. Therefore, when a user reports an error while filling in a complex form, or clicks submit at the end, the system finds many errors after checking and gives an error prompt, which only informs that the filling error needs to be corrected, but does not tell the type of error; and the user needs to rework and modify, which greatly wastes time. Therefore, although the employee has completed the filling, the content quality and overall requirements may not meet the expectations of the manager. Therefore, when the manager is reviewing, unqualified content will be rejected, and the employee needs to fill it out again based on the approval opinion. This has caused great distress to both employees and managers. Summary of the invention

[0004] In view of this, an object of the present invention is to provide a sales target filling auxiliary method, device, equipment and readable storage medium to improve the accuracy and efficiency of target filling.

[0005] In a first aspect, a sales target filling auxiliary method is provided, the method comprising:

[0006] Detect and obtain the filled-in sales target name information on the sales target filling page;

[0007] Extracting keywords from the sales target name information, where the keywords at least include region, product name, and time;

[0008] Search the target historical sales data and target filling requirements in the pre-built sales data knowledge base based on keywords. The target historical sales data at least includes sales amount and sales strategy;

[0009] Based on historical sales data, a pre-trained large model is used to generate suggestions and reference cases for filling in sales target name information;

[0010] The filling requirements, filling suggestions and reference cases are presented on the target filling page to assist in filling in the sales target elements. The sales target elements include at least sales amount, advertising investment, market share and promotion methods.

[0011] Optionally, searching for target historical sales data in a pre-built sales data knowledge base based on keywords includes:

[0012] Generate a first semantic vector and a first business feature vector based on the keywords respectively;

[0013] Use SQL data query statements to search for historical sales data containing any keyword in the pre-built database;

[0014] Preprocess the historical sales data to obtain structured text of sales data;

[0015] Generate a second semantic vector and a second business feature vector based on the sales data structured text respectively;

[0016] The target historical sales data is determined based on the similarity between the first semantic vector and the second semantic vector and the similarity between the first business feature vector and the second business feature vector.

[0017] Optionally, generating the first service feature vector based on the keyword includes:

[0018] Determine product type weight based on product name;

[0019] Determine market heat weights based on region and time;

[0020] Determining a time sensitivity weight based on time;

[0021] A first business feature vector is generated based on the product type weight, the market popularity weight, and the time sensitivity weight.

[0022] Optionally, determining the target historical sales data based on the similarity between the first semantic vector and the second semantic vector and the similarity between the first business feature vector and the second business feature vector includes:

[0023] Calculate the semantic similarity between the first semantic vector and the second semantic vector based on cosine similarity;

[0024] The business similarity between the first business feature vector and the second business feature vector is calculated based on the weighted Euclidean distance algorithm. The calculation formula is:

[0025]

[0026] The comprehensive similarity is calculated based on semantic similarity and business similarity. The calculation formula is:

[0027]

[0028] in, is the cosine similarity of the semantic vectors; is the similarity of the business feature vector; is a weight parameter used to balance the importance of semantic similarity and business similarity; Indicates the product type weight, represents the market heat weight, represents the time sensitivity weight; , Both represent business feature vectors; , Both represent semantic vectors; A represents the first and B represents the second;

[0029] The comprehensive similarity is compared with a preset similarity threshold, and if it exceeds the preset similarity threshold, the historical sales data corresponding to the second semantic vector and the second business feature vector are determined as target historical sales data.

[0030] Optionally, the method further comprises:

[0031] The filled-in sales amount is input into the pre-trained error detection model for detection to obtain the error category. The training process of the error detection model includes:

[0032] Obtain a historical sales amount dataset marked with correct and incorrect labels; the incorrect labels at least include an incorrect label for the amount data format, an incorrect label for the amount exceeding the range, and an incorrect label for the amount being a negative number;

[0033] Inputting the historical sales amount data set marked with correct and incorrect labels into the error detection model to be trained for iterative training to obtain an optimized error detection model;

[0034] Generate error prompt information according to the error category;

[0035] Present error message on the sales target filling page.

[0036] Optionally, after the sales target elements are filled in, the method further includes:

[0037] Inputting the description text of the sales target element into the hidden Markov model to convert the description text of the sales target element into structured element text sequence features;

[0038] Input the structured element text sequence features into the multi-layer perceptron to obtain the quality score of the target filling;

[0039] Compare the quality score with a preset score;

[0040] If the preset score is exceeded, the submit button of the sales target filling page is set to an operable state, otherwise it is set to an inoperable state.

[0041] Optionally, the multi-layer perceptron includes an input layer, a first hidden layer, a second hidden layer and an output layer, the second hidden layer is a hidden layer related to the business, and the structured element text sequence features are input into the multi-layer perceptron to obtain the quality score of the target filling, including:

[0042] The structured element text sequence features are input into the first hidden layer through the input layer; the first result is obtained after being processed by the first hidden layer , and then the second result is obtained after the hidden layer processing related to the business , output quality score after output layer , the calculation formula is:

[0043]

[0044]

[0045]

[0046] in, , , is the weight matrix, , , is the bias vector, is the activation function.

[0047] In a second aspect, a sales target filling auxiliary device is provided, the device comprising:

[0048] A detection unit, used to detect and obtain the filled-in sales target name information on the sales target filling page;

[0049] An extraction unit, used to extract keywords from the sales target name information, where the keywords at least include region, product name, and time;

[0050] A search unit, used to search for target historical sales data and target filling requirements in a pre-built sales data knowledge base based on keywords, where the target historical sales data at least includes sales amount and sales strategy;

[0051] A generation unit, used to generate filling suggestions and reference cases for sales target name information through a pre-trained large model based on historical sales data;

[0052] The presentation unit is used to present filling requirements, filling suggestions and reference cases on the target filling page to assist in filling in sales target elements. The sales target elements at least include sales amount, advertising investment, market share and promotion method.

[0053] In a third aspect, an electronic device is provided, comprising a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other via the communication bus;

[0054] Memory, used to store computer programs;

[0055] The processor is used to implement any method step described in the first aspect when executing the program stored in the memory.

[0056] According to a fourth aspect, a computer-readable storage medium is provided, wherein a computer program is stored in the computer-readable storage medium, and when the computer program is executed by a processor, any method step described in the first aspect is implemented.

[0057] An embodiment of the present invention provides a sales target filling auxiliary method, device, equipment and readable storage medium, which detects and obtains the filled-in sales target name information on the sales target filling page; extracts keywords in the sales target name information; searches for the target historical sales data and target filling requirements in a pre-built sales data knowledge base based on the keywords; generates filling suggestions and reference cases for the sales target name information through a pre-trained large model based on historical sales data; presents the filling requirements, filling suggestions and reference cases on the target filling page to assist in filling in the sales target elements. The present invention can give filling requirements, filling suggestions and reference cases based on the input sales target name information, and help users understand the filling rules, purpose and reasonable formulation scope of sales target elements from multi-faceted information, which significantly improves the work efficiency and accuracy in the target filling process.

[0058] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0060] Figure 1 A flow chart showing a sales target filling auxiliary method provided by an embodiment of the present invention;

[0061] Figure 2 A schematic diagram showing the structure of a sales target filling auxiliary device provided by an embodiment of the present invention is shown;

[0062] Figure 3 A schematic structural diagram of an electronic device provided by an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0063] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. The components of the embodiments of the present invention generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making creative work belong to the scope of protection of the present invention.

[0064] Considering that in the prior art, when filling in complex forms, only brief information or obscure prompts are provided, and these contents are neither intuitive nor comprehensive, resulting in filling errors. Based on this, an embodiment of the present invention provides a sales target filling auxiliary method and device, which is described below through an embodiment.

[0065] The present invention provides a sales target filling auxiliary method, which is applied to the server side, such as Figure 1 Said method comprises the following steps:

[0066] Step S101: Detect and obtain the filled-in sales target name information on the sales target filling page.

[0067] In this step, if the user enters the sales target filling page, when the user hovers the mouse or clicks the input box, the preset input event is triggered and the input sales target name information is obtained.

[0068] Step S102: extracting keywords from the sales target name information, where the keywords at least include region, product name, and time.

[0069] In an example, the sales target name information is, for example, "sales amount of XX product in North China in the second quarter"; then the keywords are North China, XX product, and second quarter.

[0070] In the embodiment of the present application, the jieba Chinese word segmentation library can be used to segment the input name information.

[0071] Step S103: Searching for target historical sales data and target filling requirements in a pre-built sales data knowledge base based on keywords, where the target historical sales data at least includes sales amount and sales strategy.

[0072] In this step, sales data related to goal setting is widely collected from various internal systems of the enterprise and external industry databases. This includes not only employees' past goal setting records and completion status, but also market dynamics, competitor information, industry standards, etc.

[0073] Requirements for filling in the target are, for example, "the target value must be greater than the minimum value, the challenge value must be greater than the target value, and the three values ​​cannot be the same."

[0074] In a specific business scenario, when salesperson Xiao Li enters "second quarter sales target", the system quickly finds relevant content from the pre-processed sales data knowledge base through keyword matching, such as historical sales trends of different products, market share changes, etc. The big model generates personalized sales strategy suggestions for Xiao Li based on this information and current market dynamics data. For example, for a new product, it is recommended to focus on expanding specific customer groups and give a reasonable range of expected sales.

[0075] Step S104: Based on historical sales data, a pre-trained large model is used to generate filling suggestions and reference cases for sales target name information.

[0076] In this step, suggestions may be filled in, for example, "Fill in the minimum value reasonably according to the environment, team status, and business resources, set a target value based on the completion of the minimum value, and set a challenge value based on the achievement of the target value."

[0077] In the embodiment of the present invention, a large model of Transformer architecture can be used, and the training process of the large model includes two stages: pre-training and fine-tuning. During pre-training, the large model learns the general patterns and knowledge of language by using the massive common texts on the Internet. In the fine-tuning stage, the enterprise's internal specific business data is used to enable the large model to adapt to the language characteristics and business requirements of the specific field of target setting.

[0078] In one example, in the scenario of project management goal planning, project manager Xiao Wang inputs the "new product development project goals", and the big model generates detailed project phase division and milestone setting suggestions, which include the key success factors, common risks and countermeasures of historical projects of the same type.

[0079] Step S105: presenting filling requirements, filling suggestions and reference cases on the target filling page to assist in filling in sales target elements, which at least include sales amount, advertising investment, market share and promotion method.

[0080] In this step, the filling requirements, filling suggestions and reference cases can be presented in the sidebar area of ​​the target filling page in an intuitive and clear manner for easy viewing by users.

[0081] Through the above embodiments, it can be understood that the present invention can provide filling requirements, filling suggestions and reference cases through a large model according to the input sales target name information, and help users understand the filling rules, purposes and reasonable formulation scope of sales target elements from multi-faceted information, which significantly improves the work efficiency and filling quality in the target filling process.

[0082] Based on the above embodiment, searching for target historical sales data in a pre-built sales data knowledge base based on keywords includes:

[0083] Step S103A: Generate a first semantic vector and a first business feature vector based on the keywords respectively.

[0084] Semantic vectors are a fixed-length numerical vector that is converted into text through an algorithm. These vectors can capture the semantic information of the text. Common methods for generating semantic vectors include word embedding, sentence embedding, etc.

[0085] Business feature vectors are feature vectors extracted based on specific business scenarios and requirements. These feature vectors usually contain various indicators and attributes related to the business and are used to support specific business decisions.

[0086] Step S103B: Use SQL data query statements to search for historical sales data containing any keyword in a pre-built database.

[0087] In the embodiment of the present invention, the historical sales data including any keyword is first preliminarily searched through keyword matching. For example, if the keyword is product A, all sales data related to the product may be found.

[0088] Step S103C: pre-process the historical sales data to obtain sales data structured text.

[0089] Since historical sales data may include some unstructured data, such as images, Word documents, PDF files, reports, articles, etc., it is convenient for subsequent data processing by converting it into structured text in a unified fixed format.

[0090] Step S103D: Generate a second semantic vector and a second business feature vector based on the sales data structured text.

[0091] It should be noted that the second semantic vector and the second service feature vector are generated in the same manner as the first semantic vector and the first service feature vector. The manner of generating the service feature vector will be described in detail in the following embodiments and will not be repeated here.

[0092] Step S103E: Determine the target historical sales data based on the similarity between the first semantic vector and the second semantic vector and the similarity between the first business feature vector and the second business feature vector.

[0093] In this step, the similarity between the first semantic vector and the second semantic vector, as well as the similarity between the first business feature vector and the second business feature vector can be calculated respectively, and then the similarities between the two can be combined to find the historical sales data with the greatest similarity as the target historical sales data. The specific method will be described in the following embodiments and will not be repeated here.

[0094] In the embodiment of the present invention, not only the traditional semantic vector is generated, but also the business-related feature vector is introduced, which can find the relevant content from the knowledge base more quickly and accurately, shorten the response time and improve the search efficiency.

[0095] Based on the above embodiment, generating the first service feature vector based on the keyword includes:

[0096] Step S103A1: Determine the product type weight according to the product name.

[0097] For example, if the keyword is XX product, the product type and its ranking can be determined according to the product name; the weight of the product type can be determined according to the ranking of the product type and the preset weight distribution rule. The preset weight distribution rule is, for example, that the weight of the top three product types is 0.8, the weight of the fourth to sixth product types is 0.6, the lower the ranking, the lower the weight, and so on. Among them, the ranking of product types can be ranked according to historical sales.

[0098] In a specific example, assuming that the X-model phone is the company's flagship product and has a high importance, the weight can be set to 0.8.

[0099] Step S103A2: Determine the market popularity weight according to the region and time.

[0100] For example, "XX District 2023 Q1 X-type mobile phone sales target", the region and time in the keywords are North China and the first quarter, then the importance of the region can be ranked according to the historical sales of the region, and the weight can be determined according to the ranking according to the weight allocation rules similar to step c. Similarly, the ranking of the quarter can be determined according to the sales situation of the quarter, and the weight can be set. The weights of the two are combined to determine the market heat weight.

[0101] In a specific example, for example, XX District is an important market for the company with high market demand, so the weight is set to 0.7. In the first quarter of 2023, the market competition is fierce, so the weight is set to 0.6. The market heat weight can be obtained by combining the two weights.

[0102] Step S103A3: Determine the time sensitivity weight based on time.

[0103] Time sensitivity refers to the importance of the sales period corresponding to that time, which can be determined based on historical sales. Based on the above example, the time sensitivity weight is determined in the same way. For example, the first quarter of 2023 is an important sales period, and the weight is set to 0.9.

[0104] Step S103A4: Generate a first business feature vector based on the product type weight, market popularity weight and time sensitivity weight.

[0105] In this step, the three weights can be combined to obtain the first service feature vector. For example, the first service feature vector is , based on the same method, the second service feature vector can be obtained .

[0106] in, Indicates the product type weight, represents the market heat weight, Represents the time sensitivity weight.

[0107] The embodiment of the present invention characterizes business features through three dimensions of product type, market popularity and time sensitivity, thereby improving the comprehensiveness and accuracy of the business feature vector.

[0108] Based on the above embodiment, determining the target historical sales data based on the similarity between the first semantic vector and the second semantic vector and the similarity between the first business feature vector and the second business feature vector includes:

[0109] Step S103E1: Calculating semantic similarity between the first semantic vector and the second semantic vector based on cosine similarity;

[0110] In the embodiment of the present invention, the formula of cosine similarity is:

[0111] (1);

[0112] in, represents the first semantic vector, represents the second semantic vector; is the dot product of vectors; and are the norms of the vectors respectively; among them, the operator " " means the product of the two, which is different from the dot product. The dot product is mainly used for operations between vectors; the product is used for operations between scalars.

[0113] In actual application scenarios, for example, when the title "Quarterly Sales Target" is input, it is converted into a vector The system extracts various texts related to sales targets from the knowledge base and converts them into vectors. By calculating the cosine similarity, we can find the text content that is most similar to the input title, such as examples of setting sales targets for past quarters, market analysis reports, etc.

[0114] Step S103E2: Calculate the service similarity between the first service feature vector and the second service feature vector based on the weighted Euclidean distance algorithm. The calculation formula is:

[0115] (2);

[0116] Step S103E3: Calculate the comprehensive similarity based on the semantic similarity and the business similarity, and the calculation formula is:

[0117] (3);

[0118] in, is the cosine similarity of the semantic vectors; is the similarity of the business feature vector; is a weight parameter used to balance the importance of semantic similarity and business similarity; Indicates the product type weight, represents the market heat weight, represents the time sensitivity weight; , Both represent business feature vectors; , Both represent semantic vectors; A represents the first and B represents the second;

[0119] Step S103E4: Compare the comprehensive similarity with a preset similarity threshold. If the similarity exceeds the preset similarity threshold, determine the historical sales data corresponding to the second semantic vector and the second business feature vector as the target historical sales data.

[0120] In the case of semantic vector matching, the embodiment of the present invention further matches the similarity of business features to make the searched historical sales data more accurate.

[0121] Based on the above embodiment, the method further includes:

[0122] Step S106: Input the filled-in sales amount into the pre-trained error detection model for detection to obtain the error category, wherein the training process of the error detection model includes:

[0123] Step S1061: Obtain a historical sales amount data set marked with correct and error labels; the error labels at least include an amount data format error label, an amount out of range error label, and an amount is a negative error label.

[0124] Step S1062: Input the historical sales amount data set marked with correct and incorrect labels into the error detection model to be trained for iterative training to obtain an optimized error detection model.

[0125] In this step, before inputting into the error detection model, some preprocessing can be performed on the data, such as data cleaning and screening, to improve the accuracy of the data.

[0126] In one example, the error detection model may employ a Naive Bayes classifier.

[0127] Step S107: Generate error prompt information according to the error category.

[0128] In an example, the error prompt information such as "the target value is less than the minimum value" can not only clearly indicate the error type, but also quickly understand the location where the error occurs.

[0129] Step S108: presenting an error prompt message on the sales target filling page.

[0130] In one example, the error prompt information may be presented in the form of a pop-up window on the sales target filling page.

[0131] Through the error detection mechanism of the embodiment of the present invention, not only can the error type of the filled-in target amount be detected, but also the position of the error can be quickly located, helping the user to quickly correct the error, thereby improving the accuracy and efficiency of error correction.

[0132] Based on the above embodiment, after the sales target elements are filled in, the method further includes:

[0133] Step S109: inputting the description text of the sales target element into the hidden Markov model to convert the description text of the sales target element into structured element text sequence features.

[0134] In one example, the HMM (Hidden Markov Model) conversion is as follows:

[0135] (4);

[0136] in, is the observation sequence, i.e., the sequence of words in the description text; are the parameters of HMM; It represents the probability of the observation sequence O occurring under the given model parameter λ; represents the initial state probability, Indicates that in the observation sequence In time t =1 is in state i The probability of Indicates from the state i Transfer to state j The probability of represents the output probability, Indicates time.

[0137] The embodiment of the present invention converts the unstructured target description text input by the employee into structured data to facilitate subsequent data analysis and processing.

[0138] Step S110: Input the structured element text sequence features into MLP (Multilayer Perceptron) to obtain the quality score of the target filling.

[0139] The process of calculating the quality score by using a multi-layer perceptron will be described in the following embodiments and will not be repeated here.

[0140] Step S111: Compare the quality score with a preset score.

[0141] Step S112: If the score exceeds the preset value, the submit button of the sales target filling page is set to an operable state, otherwise it is set to an inoperable state.

[0142] In an example, for example, the preset score is 80 points. If it exceeds 80 points, it means that the quality of the filling meets the requirements and can be submitted. If it is less than 80 points, it means that the quality of the filling does not meet the requirements and needs to be readjusted. For example, not all sales target elements are filled in, that is, the completeness of the filling does not meet the standards.

[0143] In a specific example, suppose employee Xiao Zhang sets a goal for the promotion of a new product:

[0144] The sales target elements filled in are: increase the market share of new product X by 10%, invest 50,000 yuan in online advertising, cover at least 10 cities with offline activities, and reach cooperation with 5 large dealers, with an expected sales volume of 1 million yuan.

[0145] In this example, after the employee completes the goal setting, the above features are extracted and a score, such as a score between 0 and 100, is output through the MLP model to evaluate the quality of the goal filling. The higher the score, the more complete and reasonable the goal filling is. For example, if all elements are filled in, the quality score will exceed 80 points. If elements are missing, such as market share, expected sales, etc., or the data of the filled elements is too large or too small, the quality score will be less than 80 points.

[0146] The embodiment of the present invention helps users to improve the content filled in by the sales target, so as to reach a higher quality standard before submitting it for approval, thereby effectively reducing the number of rework and modification times.

[0147] Based on the above embodiment, the multi-layer perceptron includes an input layer, a first hidden layer, a second hidden layer and an output layer, the second hidden layer is a hidden layer related to the business, and the structured element text sequence features are input into the multi-layer perceptron to obtain the target filling quality score including:

[0148] The structured element text sequence features are input into the first hidden layer through the input layer; the first result is obtained after being processed by the first hidden layer , and then the second result is obtained after the hidden layer processing related to the business , output quality score after output layer , the calculation formula is:

[0149] (5);

[0150] (6);

[0151] (7);

[0152] in, , , is the weight matrix, , , is the bias vector, is the activation function.

[0153] Among them, in the embodiment of the present invention, , , The weights can be automatically adjusted according to different business scenarios and corporate focus, making the scores more in line with actual conditions, providing employees with more targeted suggestions, and improving adaptability and flexibility.

[0154] The embodiment of the present invention adds a business-related hidden layer to the MLP. For example, for sales targets, a hidden layer is added to specifically process features related to sales data, such as sales volume, market share, etc. By using specially designed hidden layers to process features in a specific field, the model can be more focused on learning these features, thereby better capturing the complex relationships between these features and improving the performance of the model in this field. And by explicitly specifying that the hidden layer is used to process specific types of input data, the internal workings of the model can be made more transparent, easier to understand and explain. If the business environment changes, such as the introduction of a new marketing strategy, these changes can be quickly responded to by adjusting or retraining specific hidden layers without redoing the entire model.

[0155] In another embodiment, when the quality score is lower than 80 points, the user will adjust the filled-in content, the server records the adjustment action, and feeds back the adjustment action and the adjusted result to the big model. The big model optimizes the model parameters based on the policy gradient algorithm to provide better filling suggestions.

[0156] The policy gradient algorithm is as follows:

[0157] (8);

[0158] in, is the gradient operator symbol; Express Find the gradient; are the parameters of the policy network; is the expected return; is the sequence of actions adjusted, is the probability distribution of action sequences generated by the policy network; is the reward obtained by the action sequence; is the number of adjustments.

[0159] In a specific example, suppose the quality score is 65 points. After the score result is fed back to Xiao Wang, if Xiao Wang increases the advertising investment, this adjustment action will be recorded by the policy gradient algorithm. Assuming that this adjustment eventually increases the MLP score to 70 points, the policy gradient algorithm will give a higher reward to this adjustment action, and will be more inclined to recommend the adjustment strategy of increasing advertising investment in subsequent similar situations; if the MLP score decreases after the adjustment, the big model may suggest Xiao Wang to reduce the difficulty of the target or seek more team support.

[0160] Based on the same inventive concept, a sales target filling auxiliary device is provided, such as Figure 2 As shown, the device comprises:

[0161] The detection unit 201 is used to detect and obtain the filled-in sales target name information on the sales target filling page;

[0162] An extraction unit 202 is used to extract keywords from the sales target name information, where the keywords at least include region, product name, and time;

[0163] A search unit 203 is used to search for target historical sales data and target filling requirements in a pre-built sales data knowledge base based on keywords, where the target historical sales data at least includes sales amount and sales strategy;

[0164] A generating unit 204, configured to generate filling suggestions and reference cases for sales target name information through a pre-trained large model based on historical sales data;

[0165] The presentation unit 205 is used to present the filling requirements, filling suggestions and reference cases in the target filling page to assist in filling in the sales target elements. The sales target elements at least include sales amount, advertising investment, market share and promotion method.

[0166] Based on the same technical concept, an embodiment of the present invention further provides an electronic device, such as Figure 3 As shown, it includes a processor 301 , a communication interface 302 , a memory 303 and a communication bus 304 , wherein the processor 301 , the communication interface 302 , and the memory 303 communicate with each other via the communication bus 304 .

[0167] Memory 303, used for storing computer programs;

[0168] The processor 301 is used to implement the steps of the sales target filling auxiliary method when executing the program stored in the memory 303.

[0169] The communication bus mentioned in the above electronic device can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. The communication bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, only one thick line is used in the figure, but it does not mean that there is only one bus or one type of bus.

[0170] The communication interface is used for communication between the above electronic device and other devices.

[0171] The memory may include a random access memory (RAM) or a non-volatile memory (NVM), such as at least one disk memory. Optionally, the memory may also be at least one storage device located away from the aforementioned processor.

[0172] The above-mentioned processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

[0173] The computer program product for assisting the method for filling in sales targets provided in an embodiment of the present invention includes a computer-readable storage medium storing program code. The instructions included in the program code can be used to execute the method described in the previous method embodiment. The specific implementation can be found in the method embodiment and will not be repeated here.

[0174] The sales target filling auxiliary device provided in the embodiment of the present invention can be specific hardware on the device or software or firmware installed on the device. The implementation principle and technical effects of the device provided in the embodiment of the present invention are the same as those of the aforementioned method embodiment. For the sake of brief description, the parts not mentioned in the device embodiment can refer to the corresponding contents in the aforementioned method embodiment. Those skilled in the art can clearly understand that for the convenience and simplicity of description, the specific working processes of the systems, devices and units described above can all refer to the corresponding processes in the aforementioned method embodiment, and will not be repeated here.

[0175] In the embodiments provided by the present invention, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some communication interfaces, and the indirect coupling or communication connection of devices or units can be electrical, mechanical or other forms.

[0176] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0177] In addition, each functional unit in the embodiment provided by the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0178] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc., which can store program codes.

[0179] It should be noted that similar numbers and letters represent similar items in the following figures. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. In addition, the terms "first", "second", "third", etc. are only used to distinguish the description and are not to be understood as indicating or implying relative importance.

[0180] Finally, it should be noted that the above-described embodiments are only specific implementations of the present invention, which are used to illustrate the technical solutions of the present invention, rather than to limit them. The protection scope of the present invention is not limited thereto. Although the present invention is described in detail with reference to the above-mentioned embodiments, ordinary technicians in the field should understand that any technician familiar with the technical field can still modify the technical solutions recorded in the above-mentioned embodiments within the technical scope disclosed by the present invention, or can easily think of changes, or make equivalent replacements for some of the technical features therein; and these modifications, changes or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention. They should all be covered within the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. A sales target filling auxiliary method, characterized in that: The method comprises: Detect and obtain the filled-in sales target name information on the sales target filling page; Extracting keywords from the sales target name information, wherein the keywords at least include region, product name, and time; Based on the keywords, searching for target historical sales data and target filling requirements in a pre-built sales data knowledge base, wherein the target historical sales data at least includes sales amount and sales strategy; Searching for target historical sales data in a pre-built sales data knowledge base based on the keyword includes: Generate a first semantic vector and a first business feature vector based on the keywords respectively; Using SQL data query statements to search for historical sales data containing any of the keywords in a pre-built database; Performing data preprocessing on the historical sales data to obtain sales data structured text; Generate a second semantic vector and a second business feature vector based on the sales data structured text respectively; Determining target historical sales data based on the similarity between the first semantic vector and the second semantic vector and the similarity between the first business feature vector and the second business feature vector; Generating a first service feature vector based on the keyword includes: Determining a product type weight based on the product name; Determine the market heat weight according to the region and time; determining a time sensitivity weight based on the time; Generate a first business feature vector based on the product type weight, the market popularity weight and the time sensitivity weight; The second semantic vector and the second service feature vector are generated in the same manner as the first semantic vector and the first service feature vector; Based on the historical sales data, generate filling suggestions and reference cases for the sales target name information through a pre-trained large model; The filling requirements, filling suggestions and reference cases are presented in the target filling page to assist in filling in sales target elements, and the sales target elements at least include sales amount, advertising investment, market share and promotion method.

2. The method according to claim 1, characterized in that The determining of the target historical sales data based on the similarity between the first semantic vector and the second semantic vector and the similarity between the first business feature vector and the second business feature vector comprises: Calculating semantic similarity between the first semantic vector and the second semantic vector based on cosine similarity; The service similarity between the first service feature vector and the second service feature vector is calculated based on the weighted Euclidean distance algorithm, and the calculation formula is: The comprehensive similarity is calculated based on the semantic similarity and the business similarity, and the calculation formula is: in, is the cosine similarity of the semantic vectors; is the similarity of the business feature vector; is a weight parameter used to balance the importance of semantic similarity and business similarity; Indicates the product type weight, represents the market heat weight, represents the time sensitivity weight; , Both represent business feature vectors; , Both represent semantic vectors; A represents the first and B represents the second; The comprehensive similarity is compared with a preset similarity threshold, and if it exceeds the preset similarity threshold, the historical sales data corresponding to the second semantic vector and the second business feature vector are determined as target historical sales data.

3. The method according to claim 1, characterized in that The method further comprises: The filled-in sales amount is input into the pre-trained error detection model for detection to obtain the error category, wherein the training process of the error detection model includes: Obtain a historical sales amount data set marked with correct and incorrect labels; the incorrect labels at least include an incorrect label for the amount data format, an incorrect label for the amount exceeding the range, and an incorrect label for the amount being a negative number; Inputting the historical sales amount data set annotated with correct and incorrect labels into the error detection model to be trained for iterative training to obtain an optimized error detection model; Generate error prompt information according to the error category; The error prompt information is presented in the sales target filling page.

4. The method according to claim 1, characterized in that: After the sales target elements are filled in, the method further includes: Inputting the description text of the sales target element into a hidden Markov model to convert the description text of the sales target element into structured element text sequence features; Inputting the structured element text sequence features into a multi-layer perceptron to obtain a quality score of the target filling; Comparing the quality score with a preset score; If the preset score is exceeded, the submit button of the sales target filling page is set to an operable state, otherwise it is set to an inoperable state.

5. The method according to claim 4, characterized in that The multi-layer perceptron includes an input layer, a first hidden layer, a second hidden layer and an output layer, wherein the second hidden layer is a hidden layer related to the business, and the step of inputting the structured element text sequence features into the multi-layer perceptron to obtain the quality score of the target filling includes: The structural element text sequence feature is input into the first hidden layer via the input layer; after being processed by the first hidden layer, a first result is obtained. , and then the second result is obtained after the hidden layer processing related to the business , output quality score after output layer , the calculation formula is: in, , , is the weight matrix, , , is the bias vector, is the activation function.

6. A sales target filling auxiliary device, characterized in that: The device comprises: A detection unit, used to detect and obtain the filled-in sales target name information on the sales target filling page; An extraction unit, used to extract keywords from the sales target name information, wherein the keywords at least include region, product name, and time; A search unit, configured to search for target historical sales data and target filling requirements in a pre-built sales data knowledge base based on the keywords, wherein the target historical sales data at least includes sales amount and sales strategy; A search unit, specifically configured to generate a first semantic vector and a first business feature vector based on the keyword; Using SQL data query statements to search for historical sales data containing any of the keywords in a pre-built database; Performing data preprocessing on the historical sales data to obtain sales data structured text; Generate a second semantic vector and a second business feature vector based on the sales data structured text respectively; Determining target historical sales data based on the similarity between the first semantic vector and the second semantic vector and the similarity between the first business feature vector and the second business feature vector; Generating a first service feature vector based on the keyword includes: Determining a product type weight based on the product name; Determine the market heat weight according to the region and time; determining a time sensitivity weight based on the time; Generate a first business feature vector based on the product type weight, the market popularity weight and the time sensitivity weight; The second semantic vector and the second service feature vector are generated in the same manner as the first semantic vector and the first service feature vector; A generating unit, configured to generate filling suggestions and reference cases for the sales target name information through a pre-trained large model based on the historical sales data; A presentation unit is used to present the filling requirements, filling suggestions and reference cases in the target filling page to assist in filling in sales target elements, and the sales target elements at least include sales amount, advertising investment, market share, and promotion method.

7. An electronic device, characterized in that: It includes a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other through the communication bus; Memory, used to store computer programs; A processor, for implementing the method steps described in any one of claims 1 to 5 when executing a program stored in a memory.

8. 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 a processor, the method steps described in any one of claims 1 to 5 are implemented.

Images