Partner digital transformation prediction model and prediction method
The partner digital transformation prediction model uses VAE, RF, and K-means models to enhance data representation and prediction, addressing manual analysis and unethical behavior, improving accuracy and decision-making.
Patent Information
- Application Number
- CN202410923683.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-10
- Publication Date
- 2024-12-03
AI Technical Summary
Existing partner digital transformation prediction methods rely heavily on manual analysis, incomplete data collection, fail to identify violations, and do not address personalized business logic needs.
A partner digital transformation prediction model utilizing a VAE, RF, and K-means model to analyze business data for transformation dimensions and unethical behavior, with data preprocessing and machine learning techniques to enhance data representation and prediction accuracy.
Accurately predicts digital transformation and identifies unethical behaviors in partners, providing decision-making support while protecting customer privacy and improving prediction accuracy.
Smart Images

Figure CN119067172A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of enterprise digital transformation prediction, and in particular to a partner digital transformation prediction model and prediction method. Background Art
[0002] At present, as the process of enterprise globalization continues to intensify, the digital transformation of partners is particularly prominent. The digital transformation of partners refers to the digital transformation and transformation process carried out by enterprises or organizations and their partners. This transformation may cover various aspects, including the update of information technology, the optimization of digital operation processes, the improvement of data management and analysis capabilities, and the application of technologies such as cloud computing and artificial intelligence. Through digital transformation, partners can collaborate, innovate and provide services more effectively, improving overall business efficiency and customer experience.
[0003] However, there are few prediction methods for partner digital transformation. Most of them are based on the digital transformation of the enterprise or organization itself. The main shortcomings are as follows:
[0004] 1. Most partners’ digital transformation technologies rely too much on manual analysis or the manual conversion of paper processes into digital ones.
[0005] 2. Conduct incomplete data analysis by only collecting partial and one-sided business data.
[0006] 3. Failure to identify violations in the digital transformation process of partners. For example, a company’s main products are sold through partners (which can be understood as agents), and some partners privately reduce prices / discounts to compete for customers.
[0007] 4. Unable to address the needs of partners for personalized business logic in digital transformation. Summary of the invention
[0008] In view of the technical defects of the existing partner digital transformation prediction methods, the present invention provides a partner digital transformation prediction model and prediction method based on artificial intelligence and machine learning theory, which can accurately predict the digital transformation of partners.
[0009] The present invention protects a partner digital transformation prediction model, characterized in that it is composed of a VAE model, an RF model and a K-means model; the collected business data is input into the VAE model to obtain a latent vector, wherein the business data includes product data, sales data, price data, market data, customer data, and user data, which are used to analyze the output variables of four transformation dimensions, including reservations, revenue, gross profit margin, transaction size, and transaction speed. In the VAE model, the network layer of the encoder and decoder consists of three hidden layers and a two-dimensional averaging layer; the Adam optimizer and the learning rate scheduler are used to train the encoder and decoder networks, and the optimization target is the sum of the loss function calculated by the binary cross entropy and the KL divergence.
[0010] The latent vectors are input into the RF model to obtain a qualitative analysis of whether the partners have undergone digital transformation; the latent vectors are input into the K-means model to obtain a qualitative analysis of whether the partners have engaged in unethical behavior;
[0011] Use historical business data to construct the training set matrix X∈R m×n And perform data preprocessing, where m represents the number of data items, n represents the dimension of each data item, and data preprocessing includes default value processing, sparse coding processing, and regularization processing; then use the training set matrix after data preprocessing to train the VAE model, output the corresponding latent vector, and then use the latent vector to train the RF model and K-means model respectively.
[0012] The present invention utilizes the intelligent perception capability of the VAE model to learn the potential representation of business data, reconstruct and enhance the data. An intermediate data proxy is output through training of the training set matrix, which is called the latent vector. A subset of the input business data is used to analyze the output variables of four transformation dimensions: 1. Booking; 2. Revenue; 3. Gross margin; 4. Deal size and Deal velocity. Each data point represents a partner (Partner_site_party_key) or in multiple partner domains (e.g. New_Buying_Center).
[0013] The booking transformation dimension includes booking refunds, annual planned bookings, and actual bookings; the revenue transformation dimension includes annual total revenue, annual total revenue (after deducting taxes), monthly total revenue, and annual total net profit revenue; the gross profit margin transformation dimension includes the total gross profit margin and the total product booking gross profit margin; the transaction scale and transaction speed transformation dimension includes price list, basic cost, basic price, shipment confirmation date, and booking date.
[0014] The present invention also protects a partner digital transformation prediction method, which performs partner digital transformation prediction based on the above-mentioned partner digital transformation prediction model.
[0015] Preferably, feature combinations are selected for the business data used to train the VAE model, and the selected feature combinations are adjusted based on the feedback from the model evaluation indicators to achieve the optimal feature combination selection; when actually making predictions on the digital transformation of partners, the optimal feature combination is selected to make predictions on the digital transformation of partners.
[0016] Preferably, a qualitative analysis of whether the partner has unethical behavior is obtained based on the K-means model, and the selected feature combination includes the discount combination and the rebate for ordering in the predetermined transformation dimension.
[0017] The present invention protects a partner digital transformation prediction device, including a memory, a processor and a bus; the bus is used to realize the connection and communication between the memory and the processor; the processor is used to execute a computer program stored in the memory; when the processor executes the computer program, each step in the above-mentioned partner digital transformation prediction method is realized.
[0018] The present invention also protects a computer-readable storage medium having a computer program stored thereon, characterized in that when the computer program is executed by a processor, each step in the above-mentioned partner digital transformation prediction method is implemented.
[0019] Beneficial effects of the present invention:
[0020] 1. The data preprocessing method that combines default value processing, sparse coding processing and regularization processing can quickly remove noise from complex data sources, thereby extracting high-value data and improving the accuracy of model prediction.
[0021] 2. Based on the idea of combining self-encoding and decoding, clustering and classification, a hybrid digital transformation prediction model is proposed to analyze the changes in the transformation factors (four transformation dimensions) of partners in the transformation process, effectively predict the digital transformation of partners, automatically identify unethical behaviors in the digital transformation process of partners, and provide important decision-making basis for the parent company.
[0022] 3. Make predictions based on business data without the need for customer source code, thus protecting the privacy of customers’ core business. BRIEF DESCRIPTION OF THE DRAWINGS
[0023] Figure 1 This is a structural diagram of the partner digital transformation prediction model proposed in the present invention and a schematic diagram of the prediction method. DETAILED DESCRIPTION
[0024] The present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. The embodiments of the present invention are provided for the purpose of illustration and description, and are not intended to be exhaustive or to limit the present invention to the disclosed forms. Many modifications and variations will be apparent to those of ordinary skill in the art. The embodiments are selected and described in order to better illustrate the principles and practical applications of the present invention, and to enable those of ordinary skill in the art to understand the present invention and thereby design various embodiments with various modifications suitable for specific uses.
[0025] Example 1
[0026] A partner digital transformation prediction model, composed of VAE model, RF model and K-means model; the collected business data is input into the VAE model to obtain a latent vector, wherein the business data includes product data, sales data, price data, market data, customer data, and user data, which are used to analyze the output variables of four transformation dimensions, including booking, revenue, gross profit margin, transaction scale and transaction speed. Referring to Table 1, the booking transformation dimension includes booking refund, annual planned booking quantity, and actual booking quantity; the revenue transformation dimension includes annual total revenue, annual total revenue (after tax deduction), monthly total revenue, and annual total net profit revenue; the gross profit margin transformation dimension includes the total gross profit margin and the total product booking gross profit margin; the transaction scale and transaction speed transformation dimensions include price list, basic cost, basic price, shipment confirmation date, and booking date.
[0027] Table 1
[0028]
[0029] The latent vectors are input into the RF model to obtain a qualitative analysis of whether the partners have undergone digital transformation; the latent vectors are input into the K-means model to obtain a qualitative analysis of whether the partners have engaged in unethical behavior.
[0030] The basic steps of the K-means model to identify whether a partner has unethical behavior are:
[0031] 1. Initialization: Randomly select two initial cluster centers, corresponding to the presence and absence of immoral behavior respectively;
[0032] 2. Assign data points: For each data point in the data set, calculate its distance to all cluster centers and assign it to the nearest cluster center;
[0033] 3. Update cluster center: calculate the mean of all data points in each cluster and use it as the new cluster center;
[0034] 4. Iteration: Repeat steps 2 and 3 until the cluster center no longer changes significantly or the maximum number of iterations is reached.
[0035] Use historical business data to construct the training set matrix X∈R m×n And perform data preprocessing, where m represents the number of data items and n represents the dimension of each data item. Figure 1 The data processing module in the data preprocessing module performs data preprocessing, including default value processing, sparse coding processing and regularization processing. The data preprocessing method combining default value processing, sparse coding processing and regularization processing can quickly remove noise from complex data sources, thereby extracting high-value data and improving the accuracy of model prediction.
[0036] The VAE model is trained using the training set matrix after data preprocessing, and the corresponding latent vector is output. The latent vector is then used to train the RF model and the K-means model respectively. A hybrid digital transformation prediction model is proposed based on the idea of combining autoencoding and decoding, clustering and classification. The changes in the transformation factors (four transformation dimensions) of partners in the transformation process are analyzed, and the digital transformation of partners is effectively predicted. Unethical behaviors in the digital transformation process of partners are automatically identified, providing important decision-making basis for the parent company's decision-making.
[0037] In the VAE model, the network layers of the encoder and decoder consist of three hidden layers and a two-dimensional averaging layer; the Adam optimizer and learning rate scheduler are used to train the encoder and decoder networks, and the optimization target is the sum of the loss function calculated by binary cross entropy and KL divergence. Table 2 gives the definition name, type and meaning of some of the main training data of the VAE model.
[0038] Table 2
[0039] Data definition name Data Types What the data means nonecuuiing_rev_2018 NUMERIC Non-recurring income in 2018 order_cnt_2017 INTEGER Number of orders received by partners in 2017 order_cnt_2018 INTEGER Number of orders received by partners in 2018 muti_partner_deal_id_aimt INTEGER The number of multi-partner transactions each partner participated in rate_of_2017_2018 FLOAT Ratio of multi-party transactions to various types of transactions in 2017-2018 rate_of_2018_2019 FLOAT Ratio of multi-party transactions to various types of transactions in 2018-2019 rate_of_2019_2020 FLOAT Ratio of multi-party transactions to all types of transactions in 2019-2020 rate_of_revenue_type_new FLOAT Percentage of New Revenue rate_of_hardware FLOAT Percentage of hardware reservations
[0040] Example 2
[0041] A partner digital transformation prediction method is provided, which predicts the partner digital transformation based on the partner digital transformation prediction model described in Example 1.
[0042] Select the feature combination of the business data for training the VAE model, and adjust the selected feature combination according to the feedback of the model evaluation indicators to achieve the optimal feature combination selection, refer to Figure 1 ; When actually predicting the digital transformation of partners, select the optimal feature combination to predict the digital transformation of partners.
[0043] For anomaly detection problems such as unethical behavior of partners, data mining technology is used to find the main indicators that affect product discounts, and then these product discount-related features are used to identify unethical behavior of partners. Through the analysis of experimental results, it can be found that one of the most important indicators that affects the existence of unethical behavior in the digital transformation of partners is the discount rate, so two attributes related to the discount rate (pre_discount_credit_sum, rebate_bookings_sum) are selected for detailed analysis, where pre_discount_credit_sum represents the discount, which is the discount that the company has determined in advance and can give to customers (that is, how much cheaper it can be for customers), and rebate_bookings_sum represents the rebate for ordering, which can be understood as a kickback.
[0044] Table 3
[0045]
[0046]
[0047] To meet the personalized needs of users, this embodiment can support the addition of customized partner transformation rules.
[0048] Example 3
[0049] A partner digital transformation prediction device comprises a memory, a processor and a bus; the bus is used to realize connection and communication between the memory and the processor; the processor is used to execute a computer program stored in the memory; when the processor executes the computer program, each step of the partner digital transformation prediction method described in Example 2 is implemented.
[0050] Example 4
[0051] A computer-readable storage medium having a computer program stored thereon, characterized in that when the computer program is executed by a processor, each step in the partner digital transformation prediction method described in Example 2 is implemented.
[0052] Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field and related fields without creative work should fall within the scope of protection of the present invention.
Claims
1. A partner digital transformation prediction model, characterized in that: It consists of VAE model, RF model and K-means model. The collected business data is input into the VAE model to obtain latent vectors, where the business data includes product data, sales data, price data, market data, customer data and user data, which are used to analyze the output variables of the four transformation dimensions, including booking, revenue, gross profit margin, transaction size and transaction speed. The latent vectors are input into the RF model to obtain a qualitative analysis of whether the partners have undergone digital transformation; the latent vectors are input into the K-means model to obtain a qualitative analysis of whether the partners have engaged in unethical behavior; Use historical business data to construct the training set matrix X∈R m×n And perform data preprocessing, where m represents the number of data and n represents the dimension of each data; use the training set matrix after data preprocessing to train the VAE model, output the corresponding latent vector, and then use the latent vector to train the RF model and K-means model respectively.
2. The partner digital transformation prediction model according to claim 1, characterized in that: In the VAE model, the network layers of the encoder and decoder consist of three hidden layers and a two-dimensional averaging layer; the Adam optimizer and learning rate scheduler are used to train the encoder and decoder networks, and the optimization target is the sum of the loss functions calculated by binary cross entropy and KL divergence.
3. The partner digital transformation prediction model according to claim 1, characterized in that: The booking transformation dimension includes booking refunds, annual planned bookings, and actual bookings; the revenue transformation dimension includes annual total revenue, annual total revenue (after deducting taxes), monthly total revenue, and annual total net profit revenue; the gross profit margin transformation dimension includes the total gross profit margin and the total product booking gross profit margin; the transaction scale and transaction speed transformation dimension includes price list, basic cost, basic price, shipment confirmation date, and booking date.
4. A method for predicting partner digital transformation, characterized in that: A partner digital transformation prediction is performed based on the partner digital transformation prediction model described in claim 1 or 2.
5. The partner digital transformation prediction method according to claim 4, characterized in that: Select feature combinations for the business data used to train the VAE model, and adjust the selected feature combinations based on the feedback from the model evaluation indicators to achieve the optimal feature combination selection; when actually making predictions on the digital transformation of partners, select the optimal feature combination to make predictions on the digital transformation of partners.
6. The partner digital transformation prediction method according to claim 5, characterized in that: Based on the K-means model, a qualitative analysis of whether the partners have unethical behaviors is obtained, and the selected feature combination includes the discount synthesis and the rebate for ordering in the predetermined transformation dimension.
7. The partner digital transformation prediction method according to claim 4, characterized in that: Data preprocessing includes default value processing, sparse coding processing and regularization processing.
8. A partner digital transformation prediction device, characterized in that: It includes a memory, a processor and a bus; the bus is used to realize the connection and communication between the memory and the processor; the processor is used to execute the computer program stored in the memory; when the processor executes the computer program, it realizes each step of the partner digital transformation prediction method described in any one of claims 4-7.
9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, each step of the partner digital transformation prediction method described in any one of claims 4 to 7 is implemented.
Citation Information
Patent Citations
Financial enterprise digital transformation effect evaluation system based on federal learning
CN117670097A
Partner analytics management tool
US20150347952A1
