Enterprise transaction information evaluation method based on big data and related device
By building an enterprise transaction information evaluation system based on the Transformer-Encoder model, combining user and product characterization models, the data limitations and real-time problems of traditional business opportunity mining methods are solved, and efficient and accurate business opportunity mining is achieved.
Patent Information
- Application Number
- CN202510236728.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-02-28
- Publication Date
- 2025-07-11
AI Technical Summary
Traditional business opportunity mining methods rely on limited data sources, single analysis dimensions, insufficient real-time performance and a lot of manual interventions, resulting in low accuracy in corporate transaction information evaluation.
By obtaining enterprise user and transaction data, using Transformer-Encoder model and deep learning technology to build an enterprise transaction information evaluation model, and combining user portraits and product representation models to conduct multi-dimensional data analysis and real-time business opportunity mining.
It improves the accuracy and real-timeness of corporate transaction information evaluation, reduces human intervention, and improves the efficiency and accuracy of business opportunity mining.
Smart Images
Figure CN120297827A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the technical fields of big data and data processing, and specifically relates to a method for evaluating enterprise transaction information based on big data and related devices. Background Art
[0002] With the rapid development of the Internet and information technology, the global data volume has shown exponential growth, and the big data era has arrived. Various enterprises, government agencies, and research organizations have accumulated a large amount of data resources, which cover multiple dimensions such as consumer behavior, market dynamics, competitor activities, and social public opinion. How to extract valuable information from the vast amount of data, especially to identify potential business opportunities, has become a key factor for enterprises to gain a competitive advantage.
[0003] Traditional business opportunity mining methods rely on expert experience, market research reports, and simple statistical analysis of historical data. These methods have the following limitations:
[0004] Limited data sources: Traditional methods usually rely on limited data sources, such as internal sales data or a small number of market research reports, and it is difficult to comprehensively reflect the dynamic changes of the market.
[0005] Single analysis dimension: Traditional methods mainly focus on quantitative analysis and cannot effectively combine multi-dimensional data (such as social media, consumer feedback, macroeconomic indicators, etc.) for comprehensive analysis, making it difficult to identify business opportunities hidden in complex data relationships.
[0006] Lack of real-time performance: Traditional market analysis and business opportunity mining are mostly carried out regularly, unable to cope with the rapidly changing market environment, and often miss the opportunity to capture business opportunities in a timely manner.
[0007] Excessive manual intervention and low efficiency: Traditional business opportunity mining methods require a large amount of manual intervention, with a complex process, long time consumption, and are easily affected by human factors, resulting in limitations in the efficiency and accuracy of business opportunity mining.
[0008] When evaluating existing enterprise transaction information (such as big data business opportunity mining), many challenges are still faced in the application process, such as low data processing efficiency, insufficient generalization ability of the analysis model, difficulty in meeting user personalized needs, data security and privacy issues, etc., resulting in low accuracy when evaluating enterprise transaction information. Summary of the Invention
[0009] An embodiment of this application provides a method for evaluating enterprise transaction information based on big data and related devices, which can improve the accuracy when evaluating enterprise transaction information.
[0010] The first aspect of the embodiment of this application provides a method for evaluating enterprise transaction information based on big data, and the method includes:
[0011] Obtain the first enterprise basic data set and the first transaction data set of enterprise users;
[0012] Obtain an enterprise transaction information evaluation model;
[0013] Use the enterprise transaction information evaluation model to conduct transaction information evaluation through the enterprise basic data set and the transaction data set, and obtain a transaction information evaluation result.
[0014] In a possible implementation manner, the obtaining of the enterprise transaction information evaluation model includes:
[0015] Obtain a transactional data set A and an enterprise basic data set B;
[0016] Preprocess the transactional data set A and the enterprise basic data set B to obtain a first data set;
[0017] Train an enterprise user portrait model through the first data set to obtain a user portrait model ECPM, and train a commodity representation model through the first data set to obtain a commodity representation model PRM;
[0018] Construct a model according to the user portrait model ECPM, the commodity representation model PRM, and the enterprise basic data set B to obtain an enterprise transaction information evaluation model.
[0019] In a possible implementation manner, the preprocessing of the transactional data set A and the enterprise basic data set B to obtain a first data set includes:
[0020] Concatenate the transactional data in the transactional data set A and the enterprise basic data in the enterprise basic data set B according to the enterprise identifier to obtain a first reference data set;
[0021] Aggregate the first sub-data set according to the enterprise identifier to obtain a second reference data set;
[0022] Aggregate the first sub-data set according to the transaction items to obtain a third reference data set;
[0023] Determine the first reference data set, the second reference data set, and the third reference data set as the first data set.
[0024] In a possible implementation manner, the training of the enterprise user portrait model through the first data set to obtain a user portrait model ECPM includes:
[0025] Read P pieces of data from the third reference data set, where the P pieces of data include data O, and data O includes n elements;
[0026] For the input data O = [o1, o2, o3,..., o n , where o i represents the i-th element of the input sequence O, randomly select a set of positions S ∈ {1, 2,..., n} as the positions of the mask;
[0027] S ~ Bernoulli(p)
[0028] where p is the probability that each position is selected as the mask position, with a value of 0.3, and Bernoulli(p) is the Bernoulli distribution. When the result is 1, the probability is p, and when the result is 0, the probability is 1 - p.
[0029] For each i in the selected positions S, perform a masking operation to replace o i with the mask token [MASK], and the formula is as follows:
[0030]
[0031] Input the Transformer-Encoder model, and the model predicts the masked elements through context elements. Specifically, for each masked position i ∈ S, the model generates a probability distribution o′ i , and the probability distribution represents the probability of all elements appearing at that position;
[0032] Calculate the loss through the first loss function and perform backpropagation. When the loss value converges, obtain the user portrait model ECPM.
[0033] In a possible implementation, the first loss function can be represented by the following formula:
[0034]
[0035] where loss is the loss value of the first loss function, o i represents the i-th element of the input sequence O, o i,id represents the id-th element after one-hot encoding of o i , with a value of 0 or 1, o′ i,id represents the id-th element of the prediction distribution, recent_weight id the id-th represents the weight, Σ is the symbol for summing over, and log(*) is the logarithmic operation.
[0036] The second aspect of the embodiments of the present application provides an enterprise transaction information evaluation device based on big data, and the device includes:
[0037] The first acquisition unit is used to acquire the first enterprise basic data set and the first transaction data set of enterprise users;
[0038] A second acquisition unit, configured to acquire an enterprise transaction information evaluation model;
[0039] An evaluation unit, configured to perform transaction information evaluation on the enterprise basic data set and the transaction data set by using the enterprise transaction information evaluation model, so as to obtain a transaction information evaluation result.
[0040] In a possible implementation manner, the second acquisition unit is specifically configured to:
[0041] Acquire a transaction data set A and an enterprise basic data set B;
[0042] Preprocess the transaction data set A and the enterprise basic data set B to obtain a first data set;
[0043] Train an enterprise user portrait model by using the first data set to obtain a user portrait model ECPM, and train a commodity representation model by using the first data set to obtain a commodity representation model PRM;
[0044] Construct a model according to the user portrait model ECPM, the commodity representation model PRM and the enterprise basic data set B to obtain an enterprise transaction information evaluation model.
[0045] In a possible implementation manner, in terms of preprocessing the transaction data set A and the enterprise basic data set B to obtain a first data set, the second acquisition unit is specifically configured to:
[0046] Concatenate the transaction data in the transaction data set A and the enterprise basic data in the enterprise basic data set B according to the enterprise identifier to obtain a first reference data set;
[0047] Aggregate the first sub-data set according to the enterprise identifier to obtain a second reference data set;
[0048] Aggregate the first sub-data set according to the transaction item to obtain a third reference data set;
[0049] Determine the first reference data set, the second reference data set and the third reference data set as the first data set.
[0050] In a possible implementation manner, in terms of training an enterprise user portrait model by using the first data set to obtain a user portrait model ECPM, the second acquisition unit is specifically configured to:
[0051] Read P pieces of data from the third reference data set, where the P pieces of data include data O, and the data O includes n elements;
[0052] For the input data O = [o1, o2, o3,..., o n , where o i represents the i-th element of the input sequence O, randomly select a set of positions S ∈ {1, 2,..., n} as the positions of the mask;
[0053] S ∼ Bernoulli(p)
[0054] where p is the probability that each position is selected as the mask position, with a value of 0.3, and Bernoulli(p) is the Bernoulli distribution. When the result is 1, the probability is p, and when the result is 0, the probability is 1 - p.
[0055] For each i in the selected positions S, perform a masking operation to replace o i with the mask token [MASK]. The formula is as follows:
[0056]
[0057] Input the Transformer-Encoder model, and the model predicts the masked elements through context elements. Specifically, for each masked position i ∈ S, the model generates a probability distribution o′ i , and the probability distribution represents the probability of all elements appearing at this position;
[0058] Calculate the loss through the first loss function and perform backpropagation. When the loss value converges, obtain the user portrait model ECPM.
[0059] In a possible implementation, the first loss function can be represented by the following formula:
[0060]
[0061] where loss is the loss value of the first loss function, o i represents the i-th element of the input sequence O, o i,id represents the id-th element after one-hot encoding of o i , with a value of 0 or 1, o′ i,id represents the id-th element of the prediction distribution, recent_weight id the id-th represents the weight, Σ is the summation symbol for traversal, and log(*) is the logarithmic operation.
[0062] A third aspect of the embodiments of the present application provides a terminal, including a processor, an input device, an output device, and a memory. The processor, the input device, the output device, and the memory are interconnected. Among them, the memory is used to store a computer program, and the computer program includes program instructions. The processor is configured to call the program instructions to execute the step instructions in the first aspect of the embodiments of the present application.
[0063] A fourth aspect of the embodiments of the present application provides a computer-readable storage medium. Among them, the computer-readable storage medium stores a computer program for electronic data exchange. Among them, the computer program enables a computer to execute some or all of the steps described in the first aspect of the embodiments of the present application.
[0064] A fifth aspect of the embodiments of the present application provides a computer program product. Among them, the computer program product includes a non-transitory computer-readable storage medium storing a computer program. The computer program is operable to enable a computer to execute some or all of the steps described in the first aspect of the embodiments of the present application. The computer program product can be a software installation package.
[0065] Implementing the embodiments of the present application has the following beneficial effects:
[0066] By obtaining the first enterprise basic data set and the first transaction data set of enterprise users, obtaining an enterprise transaction information evaluation model, and using the enterprise transaction information evaluation model to evaluate transaction information through the enterprise basic data set and the transaction data set, a transaction information evaluation result is obtained. Therefore, it is possible to evaluate the transaction information evaluation result through the enterprise transaction information evaluation model, and the accuracy of enterprise transaction information evaluation can be improved. BRIEF DESCRIPTION OF THE DRAWINGS
[0067] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.
[0068] Figure 1 FIG. is a schematic flowchart of a method for evaluating enterprise transaction information based on big data provided by the embodiments of the present application;
[0069] Figure 2 FIG. is a specific training process diagram of the enterprise user portrait model ECPM and the commodity representation model PRM in the present application;
[0070] Figure 3 FIG. is a specific training process diagram of ResDeepFM in the present application;
[0071] Figure 4 It is a schematic structural diagram of a terminal provided by an embodiment of the present application;
[0072] Figure 5 An embodiment of the present application provides a schematic structural diagram of an enterprise transaction information evaluation device based on big data. Detailed implementation manners
[0073] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present application.
[0074] The terms "first", "second", etc. in the description and claims of the present application and the above accompanying drawings are used to distinguish different objects, rather than to describe a specific order. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device that includes a series of steps or units is not limited to the listed steps or units, but optionally further includes steps or units not listed, or optionally further includes other steps or units inherent to these processes, methods, products or devices.
[0075] Referring to "embodiment" in the present application means that the specific features, structures or characteristics described in conjunction with the embodiment may be included in at least one embodiment of the present application. The phrase appears in various positions in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment mutually exclusive with other embodiments. Those skilled in the art explicitly and implicitly understand that the embodiments described in the present application can be combined with other embodiments.
[0076] Please refer to Figure 1 , Figure 1 which is a schematic flow chart of a method for evaluating enterprise transaction information based on big data provided by an embodiment of the present application. As Figure 1 shown, the method includes:
[0077] 101. Obtain a first set of enterprise basic data and a first set of transaction data of an enterprise user.
[0078] Specifically, extract all enterprise transaction form data and enterprise user information tables from the business database; extract enterprise user information, including but not limited to basic information such as enterprise identification (ID), contact information, registration time, industry classification, etc., to obtain the first enterprise basic data set. Extract data from the enterprise user transaction data table, including but not limited to enterprise ID, information about traded items (such as item name, quantity, price, etc.), and key information such as transaction time, to obtain the first transaction data set.
[0079] Check the missing values and outliers in the first enterprise basic data set and the first transaction data set item by item. If missing values are found, fill them with the mean of the feature; if outliers are found, directly delete the relevant data rows to ensure the accuracy and consistency of the data; convert date-type data to timestamp format and extract key date information from it, such as year, month, day, etc., for subsequent analysis and modeling; perform label encoding on categorical features. When encoding, a general encoding method can be used for encoding processing.
[0080] 102. Obtain an enterprise transaction information evaluation model.
[0081] Specifically, a method for obtaining an enterprise transaction information evaluation model includes:
[0082] A1. Obtain a transactional data set A and an enterprise basic data set B;
[0083] A2. Preprocess the transactional data set A and the enterprise basic data set B to obtain a first data set;
[0084] A3. Train an enterprise user portrait model through the first data set to obtain a user portrait model ECPM, and train a commodity representation model through the first data set to obtain a commodity representation model PRM;
[0085] A4. Construct a model based on the user portrait model ECPM, the commodity representation model PRM, and the enterprise basic data set B to obtain an enterprise transaction information evaluation model.
[0086] The method for obtaining the transactional data set A and the enterprise basic data set B can refer to the method for obtaining the transactional data set and the enterprise basic data set in the foregoing embodiments, which will not be elaborated here.
[0087] In a possible implementation manner, the preprocessing of the transactional data set A and the enterprise basic data set B to obtain a first data set includes:
[0088] B1. Concatenate the transactional data in the transactional data set A and the enterprise basic data in the enterprise basic data set B according to the enterprise identification to obtain a first reference data set;
[0089] B2. Aggregate the first sub - dataset according to the enterprise identifier to obtain a second reference dataset;
[0090] B3. Aggregate the first sub - dataset according to the traded items to obtain a third reference dataset;
[0091] B4. Determine the first reference dataset, the second reference dataset, and the third reference dataset as the first dataset.
[0092] In a possible implementation, training the enterprise user portrait model through the first dataset to obtain the user portrait model ECPM includes:
[0093] C1. Read P pieces of data from the third reference dataset. The P pieces of data include data O, and data O includes n elements;
[0094] C2. For the input data O = [o1, o2, o3,..., o n , where o i represents the i - th element of the input sequence O, randomly select a set of positions S ∈ {1, 2,..., n} as the positions of the mask;
[0095] S ∼ Bernoulli(p)
[0096] where p is the probability that each position is selected as the mask position, with a value of 0.3, and Bernoulli(p) is the Bernoulli distribution. When the result is 1, the probability is p, and when the result is 0, the probability is 1 - p.
[0097] C3. For each i in the selected positions S, perform a masking operation to replace o i with the mask token [MASK]. The formula is as follows:
[0098]
[0099] C4. Input the Transformer - Encoder model. The model predicts the masked elements through context elements. For each masked position i ∈ S, the model generates a probability distribution o′ i , and the probability distribution represents the probability of all elements appearing at that position;
[0100] Calculate the loss through the first loss function and perform backpropagation. When the loss value converges, obtain the user portrait model ECPM.
[0101] In a possible implementation, the first loss function can be represented by the following formula:
[0102]
[0103] Among them, loss is the loss value of the first loss function, o i represents the i-th element of the input sequence O, o i,id represents o i the id-th element after one-hot encoding, with a value of 0 or 1, o′ i,id represents the id-th element of the predicted distribution, recent_weight id The id-th represents the weight, Σ is the summation symbol for traversal, and log(*) is the logarithmic operation.
[0104] 103. Use the enterprise transaction information evaluation model to evaluate transaction information through the first enterprise basic data set and the first transaction data set, and obtain the transaction information evaluation result.
[0105] Among them, the transaction information evaluation result can be the evaluation result when conducting business opportunity mining, etc.
[0106] In this example, by obtaining the first enterprise basic data set and the first transaction data set of enterprise users, obtaining the enterprise transaction information evaluation model, and using the enterprise transaction information evaluation model to evaluate transaction information through the enterprise basic data set and the transaction data set, the transaction information evaluation result is obtained. Therefore, it is possible to evaluate the transaction information evaluation result through the enterprise transaction information evaluation model, which can improve the accuracy of enterprise transaction information evaluation.
[0107] In a possible implementation manner, the embodiment of the present application provides another method for evaluating enterprise transaction information based on big data, which is specifically as follows:
[0108] 201. Collect the information data and transaction data of enterprise users to generate a transaction dataset A and an enterprise basic dataset B. The steps include:
[0109] Step 1. Extract all enterprise transaction form data and enterprise user information tables from the business database.
[0110] Step 2. Extract enterprise user information, including but not limited to basic information such as enterprise ID, contact information, registration time, industry classification, etc. to obtain the enterprise basic dataset B. Extract data from the enterprise user transaction data table, including but not limited to enterprise ID, information about transaction items (such as item name, quantity, price, etc.) and key information such as transaction time to obtain the transaction dataset A.
[0111] Step 3. Check the missing values and outliers in the transactional dataset A and the enterprise basic dataset B one by one. If missing values are found, fill them with the mean value of the feature; if outliers are found, directly delete the relevant data rows to ensure data accuracy and consistency; convert date-type data to timestamp format and extract key date information such as year, month, and day for subsequent analysis and modeling; perform label encoding on categorical features.
[0112] 202. Perform data preprocessing on the transactional dataset A and the enterprise basic dataset B and store them in the data center. The steps include:
[0113] Step 1. Concatenate the transactional dataset A and the enterprise basic dataset B according to the enterprise ID to form the first reference dataset C of "seller - seller information - buyer - buyer information - transaction item - transaction information", and sort it according to the transaction time.
[0114] Step 2. Aggregate C according to the enterprise ID to obtain the "enterprise - transaction item sequence data" (the second reference dataset I), and aggregate C according to the transaction item to obtain the "transaction item - enterprise sequence data" (the third reference dataset) H.
[0115] Step 3. Store A, B, C, H, and I in the MySql data center. Among them, MySql is a computer program that can run on a computer-readable storage medium, and data can be stored on the computer-readable storage medium through the program.
[0116] 203. Train the enterprise user portrait model ECPM and the commodity representation model PRM. Taking the training of the enterprise user portrait model ECPM as an example, its training process is as follows Figure 2 Similarly, the enterprise user portrait model ECPM can be trained, and its steps include:
[0117] S1. Read the dataset H, a total of P data, and take one piece of data O as an example. O has a total of n elements.
[0118] S2. Use the Transformer-Encoder structure and adopt the pre-training method of masked language model (MLM) without replacement to obtain the enterprise user portrait model ECPM. The steps include:
[0119] S3. Input the data O = [o1, o2, o3,..., o n , where o i represents the i-th element of the input sequence o, and randomly select a set of positions S ∈ {1, 2,..., n} as the masked positions.
[0120] S ∼ Bernoulli(p)
[0121] Among them, p is the probability that each position is selected as the masked position, with a size of 0.3. Bernoulli(p) is the Bernoulli distribution. When the result is 1, the probability is p, and when the result is 0, the probability is 1 - p.
[0122] For each i in the selected position S, perform a masking operation on o i and replace it with the masked token [MASK]. The formula is as follows:
[0123]
[0124] S4. Input the Transformer-Encoder model, and the model predicts the masked elements through context elements. Specifically, for each masked position i ∈ S, the model generates a probability distribution o i i , and the probability distribution represents the probability of all elements appearing at that position.
[0125] S5. Calculate the loss and perform backpropagation. When calculating the loss, due to the strong timeliness of business opportunity mining, the loss function uses the recent weighted cross-entropy. The calculation process is as follows:
[0126] S51. Take out the dataset C, count the number of occurrences of all enterprises respectively to obtain the "enterprise ID - number of occurrences" data T; take the data of the most recent 3 months from the dataset C to get C′, and count the number of occurrences of all enterprises that appear in C′ respectively to obtain the "enterprise ID - number of occurrences" data T′
[0127] S52. Calculate the recent weight Recent_weight:
[0128]
[0129] Among them, id is one of all enterprise IDs, Σ represents traversing all ids, performing the operation of calculating weights for each id, and finally splicing all the results to obtain Recent_weight.
[0130] S53. Calculate the loss. The formula is as follows:
[0131]
[0132] Among them, o i represents the i-th element of the input sequence O, o i,id represents o i after one-hot encoding, the value of the id-th element is 0 or 1, o i i ,id represents the id-th element of the prediction distribution, recent_weight idRepresents the id-th weight, and Σ is the summation symbol.
[0133] By assigning higher weights to the most recent data, the recent weighted cross-entropy effectively reduces the interference of outdated information on model predictions, enabling the model to pay more attention to and value the latest market dynamics and user behaviors, thereby better representing products and enterprises and improving the accuracy of mining current business opportunities.
[0134] 204. Construct a multi-dimensional enterprise portrait through ECPM and PRM, and jointly train the deep neural network model ResDeepFM (enterprise transaction information evaluation model) with enterprise basic data. Its training process is as Figure 3 shown, and its steps include:
[0135] Step 41. Input the enterprise ID and transaction item ID into the models EPCM and PRM respectively to obtain the representation information of each enterprise and each item, and splice them to obtain Embedding-feature. Extract the basic information from the dataset C to form the dataset X.
[0136] Step 42. Splice Embedding_feature and X to obtain E, and input it into Residual connections-DNN. The calculation formula is as follows:
[0137] E = concat(Embedding_feature, X)
[0138] E′ = linear(relu(linear(E))) + E
[0139] y DNN = linear(E′)
[0140] where linear represents the linear layer, and y DNN is the output of Residual connections-DNN.
[0141] Step 43. Input X into the FM part. The calculation formula is as follows:
[0142] X = [x0, x1, x2,..., x l
[0143]
[0144] y = (α * y FM + β * y DNN )
[0145] where l is the number of features of the dataset X, j and k are used to mark the j-th and k-th features of X, ω0 is the global bias term, and ω jis the weight associated with the j-th feature x j , v j , v k is the embedding of feature x j , x k , used to capture the interactions between features. <v j , v k > represents the inner product of the embeddings of feature x j x k . y is the final prediction result of the model. α and β are two learnable parameters with initial values of 0.5 and 0.5
[0146] Step 44. Perform backpropagation and use gradient descent to update the parameters until the enterprise transaction information evaluation model converges.
[0147] Residual connections allow information to be transmitted more directly in the network, thus accelerating the training convergence speed of the model. By reducing the loss of information between layers, the network can learn effective feature representations faster, thereby improving the training efficiency. At the same time, residual connections allow the network to learn more complex feature combinations because it can not only capture the layer-by-layer feature changes but also retain the original input information. This combination and fusion of information enhance the expressive power of the network.
[0148] 205. Run the system on a specific device. The system includes a data processing module that can preprocess the collected data and store it in the data center; a business opportunity mining module that uses the enterprise transaction information evaluation model for prediction; a recommendation module that displays and pushes feasible business opportunities based on the prediction results; a user feedback module that allows users to evaluate and provide feedback on the pushed business opportunities; a security and privacy protection module that equips the device with functions such as data encryption, access control, and log auditing to ensure the security of user data and business information.
[0149] Therefore, through data cleaning, preprocessing, and feature extraction techniques, the quality and consistency of the input data are ensured, thereby improving the accuracy and reliability of business opportunity mining.
[0150] Through advanced natural language processing and deep learning techniques, efficient embeddings of products, enterprise portraits, etc. are performed. This embedding method can capture the complex relationships and potential interaction patterns between items, helping to discover potential associated business opportunities.
[0151] Using deep learning models, especially the multi-layer structure of neural networks, this automated feature learning reduces the complexity and uncertainty of human intervention, ensuring that the model can extract the most relevant features from the data, improving the accuracy and effectiveness of business opportunity mining while reducing the labor cost and time cost in the model development process.
[0152] It adopts advanced big data analysis algorithms and deep learning models, which can automatically identify potential business opportunities from massive data. Through comprehensive analysis of multi-dimensional data such as market trends, user behavior, and competitive landscape, it accurately locates the growth points and gaps in the market, providing actionable business opportunity suggestions for enterprises.
[0153] It supports real-time data update and analysis, can promptly capture dynamic changes in the market, and helps enterprises quickly respond to market fluctuations. The real-time advantage ensures that enterprises do not miss any potential business opportunities, significantly improving the timeliness of decision-making.
[0154] Consistent with the above embodiments, please refer to Figure 4 , Figure 4 which is a schematic structural diagram of a terminal provided by an embodiment of the present application. As Figure 4 shown, it includes a processor, an input device, an output device, and a memory. The processor, input device, output device, and memory are interconnected. Among them, the memory is used to store computer programs, and the computer programs include program instructions. The processor is configured to call the program instructions, and the above programs include instructions for performing the following steps;
[0155] Obtain a first enterprise basic data set and a first transaction data set of enterprise users;
[0156] Obtain an enterprise transaction information evaluation model;
[0157] Use the enterprise transaction information evaluation model to evaluate transaction information through the enterprise basic data set and the transaction data set, and obtain a transaction information evaluation result.
[0158] The above mainly introduces the solution of the embodiment of the present application from the perspective of the execution process on the method side. It can be understood that in order for the terminal to implement the above functions, it includes the corresponding hardware structures and / or software modules for executing each function. Those skilled in the art should easily realize that, combined with the units and algorithm steps of each example described in the embodiments provided in this article, the present application can be implemented in the form of hardware or a combination of hardware and computer software. Whether a certain function is executed in the way of hardware or computer software driving hardware depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present application.
[0159] The embodiments of the present application may divide the functions of the terminal according to the above method examples. For example, each function unit may be divided corresponding to each function, or two or more functions may be integrated into one processing unit. The above integrated unit may be implemented in the form of hardware or in the form of a software function unit. It should be noted that the division of units in the embodiments of the present application is illustrative, only a logical function division, and there may be other division methods in actual implementation.
[0160] Consistent with the above, please refer to Figure 5 , Figure 5 which is a schematic structural diagram of an enterprise transaction information evaluation device based on big data provided by the embodiments of the present application. As Figure 5 shown, the device includes:
[0161] A first acquisition unit 501, configured to acquire a first enterprise basic data set and a first transaction data set of an enterprise user;
[0162] A second acquisition unit 502, configured to acquire an enterprise transaction information evaluation model;
[0163] An evaluation unit 503, configured to perform transaction information evaluation through the enterprise basic data set and the transaction data set by using the enterprise transaction information evaluation model, and obtain a transaction information evaluation result.
[0164] In a possible implementation manner, the second acquisition unit 502 is specifically configured to:
[0165] Acquire a transaction data set A and an enterprise basic data set B;
[0166] Perform preprocessing on the transaction data set A and the enterprise basic data set B to obtain a first data set;
[0167] Train an enterprise user portrait model through the first data set to obtain a user portrait model ECPM, and train a commodity representation model through the first data set to obtain a commodity representation model PRM;
[0168] Construct a model according to the user portrait model ECPM, the commodity representation model PRM and the enterprise basic data set B to obtain an enterprise transaction information evaluation model.
[0169] In a possible implementation manner, in terms of performing preprocessing on the transaction data set A and the enterprise basic data set B to obtain a first data set, the second acquisition unit 502 is specifically configured to:
[0170] Splice the transaction data in the transaction data set A and the enterprise basic data in the enterprise basic data set B according to the enterprise identifier to obtain a first reference data set;
[0171] Aggregate the first sub-dataset according to the enterprise identifier to obtain a second reference dataset;
[0172] Aggregate the first sub-dataset according to the traded items to obtain a third reference dataset;
[0173] Determine the first reference dataset, the second reference dataset, and the third reference dataset as the first dataset.
[0174] In a possible implementation manner, in the aspect of training the enterprise user portrait model through the first dataset to obtain the user portrait model ECPM, the second acquisition unit 502 is specifically configured to:
[0175] Read P pieces of data from the third reference dataset, where the P pieces of data include data O, and data O includes n elements;
[0176] For the input data O = [o1, o2, o3,..., o n , where o i represents the i-th element of the input sequence O, randomly select a set of positions S ∈ {1, 2,..., n} as the positions of the mask;
[0177] S ∼ Bernoulli(p)
[0178] where p is the probability that each position is selected as the position of the mask, with a size of 0.3, and Bernoulli(p) is the Bernoulli distribution. When the result is 1, the probability is p, and when the result is 0, the probability is 1 - p.
[0179] For each i in the selected positions S, perform a masking operation to replace o i with the mask token [MASK], and the formula is as follows:
[0180]
[0181] Input the Transformer-Encoder model, and the model predicts the masked elements through the context elements. Specifically, for each masked position i ∈ S, the model generates a probability distribution o′ i , and the probability distribution represents the probability of all elements appearing at this position;
[0182] Calculate the loss through the first loss function and perform backpropagation. When the loss value converges, obtain the user portrait model ECPM.
[0183] In a possible implementation manner, the first loss function can be represented by the following formula:
[0184]
[0185] where loss is the loss value of the first loss function, o i represents the i-th element of the input sequence O, o i,id represents o i the id-th element after one-hot encoding, with a value of 0 or 1, o′ i,id represents the id-th element of the predicted distribution, recent_weight id the id-th represents the weight, Σ is the summation symbol for traversal, and log(*) is the logarithmic operation.
[0186] The embodiments of the present application also provide a computer storage medium. Among them, this computer storage medium stores a computer program for electronic data exchange, and this computer program enables a computer to execute some or all of the steps of any one of the enterprise transaction information evaluation methods based on big data as recorded in the above method embodiments.
[0187] The embodiments of the present application also provide a computer program product. The computer program product includes a non-transitory computer-readable storage medium storing a computer program, and this computer program enables a computer to execute some or all of the steps of any one of the enterprise transaction information evaluation methods based on big data as recorded in the above method embodiments.
[0188] It should be noted that for the foregoing method embodiments, for the sake of simple description, they are all expressed as a series of action combinations. However, those skilled in the art should know that the present application is not limited by the described action sequence, because according to the present application, certain steps can be adopted in other sequences or simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily essential to the present application.
[0189] In the above embodiments, the descriptions of the respective embodiments have their own emphases. For the parts not detailed in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.
[0190] In several embodiments provided by the present application, it should be understood that the disclosed device can be implemented in other ways. For example, the device embodiments described above are only illustrative. For example, the division of the units is only a logical function division. In actual implementation, there can be other division methods. 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, the displayed or discussed mutual coupling or direct coupling or communication connection can be through some interfaces, and the indirect coupling or communication connection of the device or unit can be in an electrical or other form.
[0191] The unit described as a separation component may or may not be physically separated. The component shown as a unit may or may not be a physical unit, that is, it may be located in one place or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.
[0192] In addition, in each embodiment of the application, each functional unit can be integrated in a processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of a software program module.
[0193] If the above-mentioned integrated unit is implemented in the form of a software program module and sold or used as an independent product, it can be stored in a computer-readable memory. Based on such an understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a memory and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in each embodiment of this application. The aforementioned memory includes: various media such as USB flash drives, read-only memories (ROM), random access memories (RAM), mobile hard disks, magnetic disks, or optical discs that can store program codes.
[0194] Those of ordinary skill in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by instructing relevant hardware through a program. This program can be stored in a computer-readable memory, and the memory can include: flash drives, read-only memories, random access memories, magnetic disks, or optical discs, etc.
[0195] The above has introduced the embodiments of this application in detail. Specific examples are used in this article to elaborate on the principle and implementation manner of this application. The description of the above embodiments is only used to help understand the method and its core idea of this application; at the same time, for those of ordinary skill in the art, according to the idea of this application, there will be changes in the specific implementation manner and application scope. In summary, the content of this specification should not be construed as a limitation to this application.
Claims
1. An enterprise transaction information evaluation method based on big data, characterized in that, The method includes: Obtaining a first set of enterprise basic data and a first set of transaction data of enterprise users; Obtaining an enterprise transaction information evaluation model; Using the enterprise transaction information evaluation model to evaluate transaction information through the enterprise basic data set and the transaction data set, and obtaining a transaction information evaluation result.
2. The enterprise transaction information evaluation method based on big data according to claim 1, wherein The obtaining of the enterprise transaction information evaluation model includes: Obtaining a transaction data set A and an enterprise basic data set B; Preprocessing the transaction data set A and the enterprise basic data set B to obtain a first data set; Training an enterprise user portrait model through the first data set to obtain a user portrait model ECPM, and training a commodity representation model through the first data set to obtain a commodity representation model PRM; Constructing a model according to the user portrait model ECPM, the commodity representation model PRM and the enterprise basic data set B to obtain an enterprise transaction information evaluation model.
3. The method for evaluating enterprise transaction information based on big data according to claim 2, wherein The preprocessing of the transaction data set A and the enterprise basic data set B to obtain a first data set includes: Concatenating the transaction data in the transaction data set A and the enterprise basic data in the enterprise basic data set B according to the enterprise identifier to obtain a first reference data set; Aggregating the first sub-data set according to the enterprise identifier to obtain a second reference data set; Aggregating the first sub-data set according to the traded items to obtain a third reference data set; Determining the first reference data set, the second reference data set and the third reference data set as the first data set.
4. The enterprise transaction information evaluation method based on big data according to claim 3, wherein, The training of the enterprise user portrait model through the first data set to obtain the user portrait model ECPM includes: Reading P pieces of data from the third reference data set, where the P pieces of data include data O, and the data O includes n elements; For the input data O = [o1, o2, o3,..., o n , where o i represents the i-th element of the input sequence O, randomly select a set of positions S ∈ {1, 2,..., n} as the positions of the mask; S~Bernoulli(p) where p is the probability that each position is selected as the masked position, with a size of 0.3, and Bernoulli(p) is the Bernoulli distribution. When the result is 1, the probability is p, and when the result is 0, the probability is 1 - p. Perform a masking operation on each i in the selected position S to replace o i with the mask token [MASK], as shown in the following formula: Input the Transformer-Encoder model. The model predicts the masked elements based on the context elements. For each masked position \(i\in S\), the model generates a probability distribution \(o\). i ′ , where the probability distribution represents the probabilities of all elements at that position. Calculating the loss through the first loss function and performing backpropagation, and obtaining the user portrait model ECPM when the loss value converges.
5. The method for evaluating enterprise transaction information based on big data according to claim 4, wherein The first loss function can be represented by the following formula: where loss is the loss value of the first loss function, o i represents the i-th element of the input sequence O, o i,id represents o i the id-th element after one-hot encoding, with a value of 0 or 1, o i ′ ,id represents the id-th element of the predicted distribution, recent_weight id the id-th represents the weight, Σ is the summation symbol for traversal, and log(*) is the logarithmic operation.
6. An enterprise transaction information evaluation device based on big data, characterized in that The device includes: A first acquisition unit for acquiring a first set of enterprise basic data and a first set of transaction data of enterprise users; A second acquisition unit for acquiring an enterprise transaction information evaluation model; An evaluation unit for using the enterprise transaction information evaluation model to evaluate transaction information through the enterprise basic data set and the transaction data set, and obtaining a transaction information evaluation result.
7. The method for evaluating enterprise transaction information based on big data according to claim 6, wherein The second acquisition unit is specifically used for: Obtaining a transaction data set A and an enterprise basic data set B; Preprocessing the transaction data set A and the enterprise basic data set B to obtain a first data set; Training an enterprise user portrait model through the first data set to obtain a user portrait model ECPM, and training a commodity representation model through the first data set to obtain a commodity representation model PRM; Model construction is performed according to the user portrait model ECPM, the product representation model PRM, and the enterprise basic data set B to obtain an enterprise transaction information evaluation model.
8. The method for evaluating enterprise transaction information based on big data according to claim 7, wherein In terms of preprocessing the transaction data set A and the enterprise basic data set B to obtain the first data set, the second acquisition unit is specifically configured to: Concatenate the transaction data in the transaction data set A and the enterprise basic data in the enterprise basic data set B according to the enterprise identifier to obtain a first reference data set; Aggregate the first sub-data set according to the enterprise identifier to obtain a second reference data set; Aggregate the first sub-data set according to the traded items to obtain a third reference data set; Determine the first reference data set, the second reference data set, and the third reference data set as the first data set.
9. A terminal, characterized in that, It includes a processor, an input device, an output device, and a memory. The processor, the input device, the output device, and the memory are interconnected. Among them, the memory is used to store a computer program, and the computer program includes program instructions. The processor is configured to call the program instructions to execute the big data-based enterprise transaction information evaluation method according to any one of claims 1-5.
10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, and the computer program includes program instructions. When the program instructions are executed by the processor, the processor is caused to execute the big data-based enterprise transaction information evaluation method according to any one of claims 1-5.