Method and device for constructing prediction model of capability portrait in business, and storage medium

Abstract

The invention discloses a prediction model construction method and device for a table capability portrait in a service and a storage medium, and relates to the technical field of artificial intelligence, and the method comprises the steps: carrying out the preprocessing of system operation data of a table system in the service, and obtaining annotation data; executing a pre-training task on the labeled data through a pre-training language model to obtain a feature extractor; and performing iterative training on the Bayesian network with uncertainty prediction based on the feature vectors extracted by the feature extractor and the annotation data to obtain a prediction model of the capability portrait in the service. According to the method, the feature vector is extracted through the pre-trained feature extractor, and the Bayesian network with uncertainty prediction is introduced for iterative training, so that dependence on knowledge and experience of service experts and quality and integrity of original data is avoided; therefore, iteration of the capability portrait in the business can be realized through the constructed prediction model, and rapid and accurate intelligent portrait analysis of various business system capabilities is realized.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to a method, device and storage medium for constructing a prediction model for business middle-end capability portrait. Background Art

[0002] The business middle platform capability portrait is a systematic description of the various capabilities possessed by the business middle platform. It can outline the capability characteristics of the business middle platform in different aspects, such as data processing capabilities, process integration capabilities, service provision capabilities, etc.

[0003] Currently, there are four main types of business middle-office capability portrait construction solutions: rule-based business middle-office capability portrait construction solution, statistical analysis-based business middle-office capability portrait construction solution, traditional machine learning model-based business middle-office capability portrait construction solution, and natural language processing (NLP)-based business middle-office capability portrait construction solution.

[0004] However, the existing business middle platform capability profiling solutions rely heavily on the knowledge and experience of business experts and the quality and integrity of original data. In real business scenarios, the various business systems of the business middle platform are constantly being updated and iterated, such as new business capabilities and some capabilities that are less frequently used by users. The existing business middle platform capability profiling solutions are difficult to adapt to changes in business systems, and there will be some situations where business system capabilities are difficult to cover. Summary of the invention

[0005] The main purpose of this application is to provide a method, device and storage medium for building a predictive model for business middle-end capability portrait, aiming to solve the technical problem that the existing business middle-end capability portrait solution is difficult to adapt to changes in the business system and there are some business system capabilities that are difficult to cover.

[0006] To achieve the above objectives, the present application proposes a method for constructing a prediction model for a business middle station capability portrait, the method comprising:

[0007] Pre-process the system operation data of the business middle-end system to obtain labeled data;

[0008] Performing a pre-training task on the labeled data through a pre-trained language model to obtain a feature extractor, wherein the pre-training task includes log prediction based on system operation logs and prediction of association between logs and business fields;

[0009] Based on the feature vector extracted by the feature extractor and the labeled data, the Bayesian network is iteratively trained to obtain a prediction model for the business middle station capability portrait, and the Bayesian network is a model with uncertainty prediction.

[0010] In one embodiment, the step of performing a pre-training task on the labeled data by using a pre-trained language model to obtain a feature extractor includes:

[0011] Load the pre-trained language model according to the pre-trained task to obtain the corresponding task prediction model;

[0012] Use cross entropy loss as the loss function;

[0013] Based on the labeled data, optimizing the task prediction model through the loss function to obtain a feature extractor;

[0014] A feature vector of the system operation data is extracted according to the feature extractor.

[0015] In one embodiment, the step of iteratively training the Bayesian network based on the feature vector extracted by the feature extractor and the labeled data to obtain a prediction model of the business middle station capability portrait includes:

[0016] According to the feature vector extracted by the feature extractor and the labeled data, the Bayesian network is trained using the lower bound of the evidence of variational inference as a minimization loss function to obtain an uncertain model;

[0017] Reasoning on the unlabeled data of the system operation data by using the uncertain model, and outputting labels and uncertainties;

[0018] According to the label and the uncertainty, the uncertain model is iteratively trained to obtain a prediction model of the business middle station capability portrait.

[0019] In one embodiment, the step of iteratively training the uncertain model according to the label and the uncertainty to obtain a prediction model of the business middle station capability portrait includes:

[0020] Manually labeling the data to be labeled according to the uncertainty to obtain manually labeled data;

[0021] Training the uncertain model based on the label and the manually annotated data to obtain an output result;

[0022] When the output result does not meet the preset stop condition, returning to execute the step of training the Bayesian network based on the feature vector extracted by the feature extractor and the labeled data, using the lower bound of the evidence of variational inference as the minimization loss function;

[0023] Until the output result reaches the preset stop condition, the model corresponding to the preset stop condition will be used as the prediction model of the business middle station capability portrait.

[0024] In one embodiment, when the preset stop condition is a performance indicator, the step of using the model corresponding to the preset stop condition as the prediction model of the business middle station capability portrait until the output result reaches the preset stop condition includes:

[0025] Determine the actual number of business middle office capability portraits in the business middle office system;

[0026] Determine the coverage and accuracy of the portrait according to the actual number and the predicted number of business middle station capability portraits in the output result;

[0027] When the portrait coverage and accuracy reach the performance indicators, the model corresponding to the performance indicators will be used as the prediction model of the business middle platform capability portrait.

[0028] In one embodiment, the step of preprocessing the system operation data of the business middle station system to obtain the labeled data includes:

[0029] Obtaining system operation data of the business middle office system, wherein the system operation data includes user behavior logs, system basic data, system functional data and operation and maintenance data;

[0030] Cleaning the user behavior log, the system basic data, the system functional data, and the operation and maintenance data to obtain cleaned data;

[0031] The cleaned data are respectively labeled with image labels to obtain labeled data.

[0032] In addition, to achieve the above purpose, the present application also proposes a prediction model construction device for business middle station capability portrait, the device comprising:

[0033] The data preprocessing module is used to preprocess the system operation data of the business middle office system to obtain labeled data;

[0034] A feature extraction module, used to perform a pre-training task on the labeled data through a pre-trained language model to obtain a feature extractor, wherein the pre-training task includes log prediction based on system operation logs and prediction of association between logs and business domains;

[0035] The model training module is used to iteratively train the Bayesian network based on the feature vector extracted by the feature extractor and the labeled data to obtain a prediction model for the business middle station capability portrait, and the Bayesian network is a model with uncertainty prediction.

[0036] In addition, to achieve the above-mentioned purpose, the present application also proposes a server, which includes: a memory, a processor, and a computer program stored on the memory and executable on the processor, and the computer program is configured to implement the steps of the method for building a predictive model of the business middle-end capability portrait as described above.

[0037] In addition, to achieve the above-mentioned purpose, the present application also proposes a storage medium, which is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by the processor, the steps of the method for constructing a predictive model of the business middle-end capability portrait as described above are implemented.

[0038] In addition, to achieve the above-mentioned purpose, the present application also provides a computer program product, which includes a computer program, and when the computer program is executed by a processor, it implements the steps of the method for constructing a predictive model of the business middle station capability portrait as described above.

[0039] One or more technical solutions proposed in this application have at least the following technical effects: this application first pre-processes the system operation data of the business middle station system to obtain labeled data; then performs pre-training tasks on the labeled data through a pre-trained language model to obtain a feature extractor, and the pre-training tasks include log prediction based on system operation logs and prediction of the relationship between logs and business fields; finally, based on the feature vectors extracted by the feature extractor and the labeled data, the Bayesian network is iteratively trained to obtain a prediction model for the business middle station capability portrait, and the Bayesian network is a model with uncertainty prediction. Since this application extracts feature vectors through a pre-trained feature extractor and introduces a Bayesian network with uncertainty prediction for iterative training, the iteration of the business middle station capability portrait can be achieved through the constructed prediction model, avoiding dependence on the knowledge and experience of business experts and the quality and integrity of original data, and realizing fast and accurate intelligent portrait analysis of various business system capabilities. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0041] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0042] Figure 1A flowchart of the first embodiment of the method for constructing a prediction model for the business platform capability portrait of this application;

[0043] Figure 2 A flowchart diagram of the second embodiment of the method for constructing a prediction model for the business platform capability portrait of this application;

[0044] Figure 3 A flowchart diagram of the third embodiment of the method for constructing a prediction model for the business platform capability portrait of this application;

[0045] Figure 4 A schematic diagram of the module structure of a device for constructing a prediction model for a business platform capability portrait in an embodiment of the present application;

[0046] Figure 5 Schematic diagram of the device structure of the hardware operating environment involved in the method for building a predictive model for the business middle station capability portrait in the embodiment of the present application.

[0047] The realization of the purpose, functional features and advantages of this application will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0048] It should be understood that the specific embodiments described herein are only used to explain the technical solutions of the present application and are not used to limit the present application.

[0049] In order to better understand the technical solution of the present application, a detailed description will be given below in conjunction with the accompanying drawings and specific implementation methods.

[0050] It should be noted that the current four business middle platform capability profiling solutions have the following shortcomings:

[0051] 1. In the rule-based business middle platform capability profiling solution, the formulation of rules usually relies on the knowledge and experience of business experts and is not flexible enough. As the business develops, maintaining and updating rules requires a lot of work, and the rule set may become large and complex and difficult to manage; this makes it difficult to adapt to rapidly changing business needs or emerging business scenarios, limiting the system's ability to adapt to new situations.

[0052] 2. The solution for building a business middle-office capability portrait based on statistical analysis focuses on using quantitative data to evaluate and describe the capabilities of the middle-office system. However, the statistical analysis of this solution is highly dependent on the quality and integrity of the data. If the data is biased or incomplete, the analysis results may be inaccurate; the statistical data of the statistical model may contain outdated data and cannot reflect the latest business changes or user behaviors in a timely manner; statistical models have specific assumptions, such as normal distribution, independence, etc.; if the data does not meet these assumptions, the model may not be applicable; it cannot adapt to the updates of the business system in a timely manner and reflect the business capabilities of each system in real time.

[0053] 3. The business middle-office capability portrait construction solution based on the traditional machine learning model is a systematic process that uses machine learning technology to analyze the business middle-office system to build a portrait that reflects the system's capabilities. However, this solution usually requires manual feature engineering, which includes feature selection, feature extraction, and feature conversion, which is a time-consuming and technically demanding process; the trained machine learning model performs well on a specific data set, but may degrade when generalized to other data sets or business scenarios; as the business environment changes, the model needs to be regularly updated and maintained, which involves retraining and adjusting model parameters.

[0054] 4. The solution for building a business middle-office capability portrait based on natural language processing mainly relies on the analysis and understanding of text data during the operation of the system to extract key information and insights about the business middle-office system. However, the terminology and abbreviations of specific industries or fields in this solution pose challenges to the construction of capability portraits, requiring additional training data or customized processing; the performance of the model is highly dependent on the quality of the data and preprocessing steps, such as word segmentation, part-of-speech tagging, named entity recognition, etc.; the text language expression of some systems is extremely personalized, and the system may find it difficult to capture the nuances and personalized needs of each text; business system text data may be high-dimensional and sparse, resulting in high difficulty in model training and interpretation; as the business environment changes, the model needs to be regularly updated and maintained to adapt to new business systems, which will involve retraining and adjusting model parameters.

[0055] In summary, in real business scenarios, various business systems are constantly updated and iterated. The above solutions may have some technical problems that cannot be covered by the business system capabilities, such as new business capabilities and some capabilities that are less frequently used by users.

[0056] Therefore, in order to solve the above problems, this application proposes a solution for business middle-end capability profiling based on autonomous learning technology. By introducing an autonomous learning framework and combining uncertain learning to achieve low-cost business capability profiling iteration, it is possible to achieve fast and accurate intelligent profiling analysis of various business system capabilities, which enables business personnel to have a clearer, more intuitive and accurate understanding of the existing capabilities of the middle-end, and can also efficiently build application scenarios and solutions.

[0057] It should be noted that the execution subject of this embodiment can be a computing service device with data preprocessing, feature extraction and model training functions, such as a personal computer, server, etc., or an electronic device that can realize the above functions, a construction device specifically used for the business middle station system to execute the prediction model construction method of the business middle station capability portrait of this application, etc. This embodiment does not limit this. The following takes the construction device as an example to illustrate this embodiment and the following embodiments.

[0058] Based on this, the embodiment of the present application provides a method for constructing a prediction model for the business middle station capability portrait, referring to Figure 1 , Figure 1 A flowchart diagram is provided for the first embodiment of the method for constructing a prediction model for the business platform capability portrait of this application.

[0059] In this embodiment, the method for constructing a prediction model for the business middle station capability portrait includes steps S10 to S30:

[0060] Step S10: pre-process the system operation data of the business middle-end system to obtain labeled data.

[0061] It should be noted that the business middle office system is a reusable service system formed by integrating and precipitating the core business capabilities of the enterprise. System operation data is the relevant data in the business middle office system that can reflect the capabilities of the business middle office system, such as system basic data, system functional data, operation and maintenance data, etc.

[0062] It should be noted that the labeled data is the data obtained after pre-processing the system operation data (for example, if there is duplicate data and useless data in the system operation data, the system operation data can be cleaned to remove these redundant data and improve the efficiency of subsequent data processing; or the data that can characterize the capabilities of the business middle platform can be labeled to obtain portrait labels belonging to various data).

[0063] In this implementation, data characterizing the capabilities of the business middle platform can be obtained by preprocessing the system operation data.

[0064] In a feasible implementation, step S10 of this embodiment may include the steps of: obtaining system operation data of the business middle-end system, wherein the system operation data includes user behavior logs, system basic data, system functional data and operation and maintenance data; performing data cleansing on the user behavior logs, the system basic data, the system functional data and the operation and maintenance data to obtain cleaned data; and labeling the cleaned data with portrait labels to obtain labeled data.

[0065] It should be noted that the user behavior log records the various operational behaviors of users when using the business middle-office system, including the time when the user accesses the system, the IP address, the pages or functions visited, and the user's evaluation and feedback on the system.

[0066] System basic data is the basic data required for the operation of the business middle-office system, including the middle-office system name, system type, system function description, system architecture description, development language and framework, etc.

[0067] System functional data refers to data related to the specific functions of the business middle-office system, including a list of system core functions, function usage frequency, function coverage, response time, network traffic throughput, failure rate, etc.

[0068] Operation and maintenance data refers to data related to the operation and maintenance of the business middle-office system, including the operation and maintenance tools and platforms used by the system, key system indicators, system alarm records and processing results, etc.

[0069] Specifically, we can first collect data related to the business middle platform system and clarify which data is crucial to understanding the capabilities of the business system, including user behavior logs, system basic data, system functional data, operation and maintenance data, etc. Then we clean the data to remove duplicate data and useless data. Then we label the cleaned data with portrait labels (i.e., labels that characterize the capabilities of the business middle platform) to obtain labeled data.

[0070] Exemplarily, for ease of understanding, for user behavior logs, duplicate records in the user behavior logs can be checked and removed, log records not related to the task (such as system internal debugging information, etc.) can be removed, and finally all timestamps can be converted into a unified format.

[0071] For system basic data, duplicate system basic data records can be checked and removed. Then, punctuation, capitalization, etc. in the text of the system basic data are normalized to ensure consistency. Then, extra spaces and special characters in the text are cleaned. Finally, the data is manually annotated to mark out the entities in the system basic data.

[0072] For system functional data, we can first merge the data. For the log content in the system functional data, we can unify the timestamp format, remove spaces and special characters in the log content, count the log level, and add the corresponding business field label to each log. For the performance data in the system functional data, we can count the frequency of function use, response time, network traffic throughput, failure rate and other information, and eliminate incomplete or invalid records.

[0073] For operation and maintenance data, we can first remove incomplete or invalid record data to obtain the system's key performance indicators (KPIs), such as CPU usage, memory usage, disk IO, network bandwidth, etc. Then collect the alarm records of the business middle office system and the corresponding processing results, including alarm type, alarm time, processing time, processing results, etc. Then remove the spaces and special characters in the log content of the operation and maintenance data. Finally, mark the relationship label between the network traffic data and the system fault alarm log.

[0074] In this implementation, the validity of the data can be further improved by performing data cleaning and labeling portrait tags on the system operation data.

[0075] Step S20: performing a pre-training task on the labeled data through a pre-trained language model to obtain a feature extractor, wherein the pre-training task includes log prediction based on system operation logs and prediction of associations between logs and business domains.

[0076] It should be noted that the pre-trained language model is a model that performs unsupervised learning on labeled data to learn the characteristics of system operation data. The pre-trained language model may include a RoBERTa model, a GPT model, etc., which is not limited in this embodiment.

[0077] It should be noted that the pre-training task is designed specifically for the system operation data, so that the pre-trained language model can understand the logical relationship and capabilities of the business middle-end system.

[0078] Exemplarily, pre-training tasks may include log prediction based on system operation logs (used for pre-training language models to understand the order and logical relationship between logs), prediction of the relationship between logs and business domains (used for pre-training language models to learn the relationship between systems and logs and capture richer semantic information), system key entity recognition tasks (used to improve the pre-training language model's understanding and reasoning capabilities of basic system information), system network fault warning tasks (used for pre-training language models to understand and reason about operation and maintenance data), etc. This embodiment does not impose any restrictions on this.

[0079] It should be noted that the feature extractor is a model of the pre-trained language model after fine-tuning the above pre-training task. At this time, the feature extractor has the ability to extract features of the system operation data and performs the final feature extraction on the system operation data.

[0080] In this implementation, two targeted pre-training tasks are designed based on relevant data such as system operation data related to each business system, namely log prediction based on system operation logs and prediction of the relationship between logs and business fields. The pre-trained language model is fine-tuned using the understanding of log time series and the understanding of log features of specific business scenarios to perform final feature extraction on the data of all parties of system operation data. At this time, the feature vector has rich system-related knowledge and can fully represent the corresponding system behavior data, supporting subsequent model training and reasoning.

[0081] In a feasible implementation, step S20 of this embodiment may include the steps of: loading a pre-trained language model according to a pre-trained task to obtain a corresponding task prediction model; using the cross entropy loss as a loss function; based on the labeled data, optimizing the task prediction model through the loss function to obtain a feature extractor; and extracting a feature vector of the system operation data according to the feature extractor.

[0082] It should be noted that the task prediction model is the corresponding model obtained after loading the pre-trained language model according to different pre-training tasks. For example, based on the system operation log, by loading the RoBERTa model, a log prediction model can be constructed; based on the system basic information data, by loading the RoBERTa model, an entity recognition model can be constructed.

[0083] It should be noted that Cross Entropy Loss is a commonly used loss function in multi-classification tasks, which is used to predict the categories and distribution of various data in the labeled data.

[0084] It should be noted that the feature vector is a numerical vector extracted from the system operation data that can describe the key characteristics and attributes of the system operation data. Through the feature extractor, representative and distinguishing features can be extracted from complex system operation data and represented as an ordered vector for subsequent model training.

[0085] For example, the RoBERTa model is used as an example for illustration, but this is not a limitation of this solution. After obtaining the labeled data, the process for the pre-training task is as follows:

[0086] 1) Log prediction task based on system operation log. Based on the system operation log, you can first load the RoBERTa model and build a log prediction model. Then use the log prediction model to determine whether the two logs are adjacent to each other, and then use the multi-class cross-entropy loss as the loss function to optimize the log prediction model. After fine-tuning the log prediction task based on the system operation log, the RoBERTa model's ability to understand and reason about system logs can be improved. Here, judging whether the two logs are adjacent to each other is to understand the order and logical relationship between the logs, which is very important for understanding the overall picture of the logs.

[0087] 2) Prediction of the relationship between logs and business domains. Based on the system operation logs, the RoBERTa model can be loaded first to build a relationship prediction model. Then the business domain to which the current log belongs is determined, and the relationship prediction model is optimized using the cross entropy loss as the loss function. Fine-tuning the task of predicting the relationship between logs and business domains also helps the RoBERTa model learn the relationship between the system and logs and capture richer semantic information.

[0088] 3) System key entity recognition. Based on the basic system information in the system operation data, the RoBERTa model can be loaded first to build an entity recognition model. Then, the key data in the basic system information is extracted, and the entity recognition model is optimized using the cross entropy loss as the loss function. Fine-tuning the system key entity recognition task also helps the RoBERTa model understand and reason about the basic system information.

[0089] 4) System network fault warning. Based on the network traffic data of the system during operation, the RoBERTa model can be loaded first to build a fault prediction model. Then, based on the current network environment and network traffic data, it is predicted whether the system will fail, and the fault prediction model is optimized using the binary cross entropy loss as the loss function. Fine-tuning the system network fault warning task also helps the RoBERTa model understand and reason about operation and maintenance data.

[0090] In this embodiment, targeted pre-training tasks are designed for different business system data so that the RoBERTa model can learn the data characteristics of different systems. Among them, log prediction based on system operation logs allows the RoBERTa model to predict the next log entry of a given log sequence, enhancing the model's understanding of log timing. The prediction of the relationship between logs and business fields allows the RoBERTa model to predict which specific business field the current log belongs to, enhancing the model's understanding of the log characteristics of specific business scenarios. Compared with the feature extraction method in the business middle-station capability portrait construction solution based on traditional machine learning models and the business middle-station capability portrait construction solution based on natural language processing, the feature vector obtained by feature extraction in this embodiment can have richer system-related knowledge, can fully represent the corresponding system behavior data, and support subsequent model training and reasoning.

[0091] Step S30: Based on the feature vector extracted by the feature extractor and the labeled data, the Bayesian network is iteratively trained to obtain a prediction model for the business middle station capability portrait, and the Bayesian network is a model with uncertainty prediction.

[0092] It should be noted that the Bayesian network is a graphical model based on probabilistic reasoning. When predicting, the Bayesian network can not only give the probability of the system operation data, but also predict the uncertainty of the prediction result.

[0093] By using the trained Bayesian network (i.e., the prediction model) to perform reasoning on the system operation data, it is possible to produce labels and uncertainties for the business middle-end capability portrait.

[0094] It should be noted that the business middle-office capability portrait is a systematic description of the various capabilities possessed by the business middle-office system. It can outline the capability characteristics of the business middle-office system in different aspects, such as data processing capabilities, process integration capabilities, and service provision capabilities.

[0095] In the iterative training process, the training can be performed by setting a preset training batch, and the prediction model can be obtained when the training batch is reached; the training can also be performed by setting a performance index, and the prediction model can be obtained when the training result reaches the performance index. This embodiment does not limit this.

[0096] In the technical solution provided in this embodiment, first, data related to the business middle station system can be collected to clarify which data is crucial to understanding the capabilities of the business system, including user behavior logs, system basic data, system functional data, operation and maintenance data, etc. Then the data is cleaned to remove duplicate data and useless data. Then the cleaned data is labeled with portrait labels (i.e., label labels that characterize the capabilities of the business middle station) to obtain labeled data. Then, for different business system data, targeted pre-training tasks are designed so that the RoBERTa model can learn the data features of different systems to obtain a feature extractor. Finally, based on the feature vectors extracted by the feature extractor and the labeled data, the Bayesian network is iteratively trained to obtain a prediction model for the business middle station capability portrait. Since this embodiment extracts feature vectors through a pre-trained feature extractor and introduces a Bayesian network with uncertainty prediction for iterative training, the iteration of the business middle station capability portrait can be achieved through the constructed prediction model, avoiding the reliance on the knowledge and experience of business experts and the quality and integrity of the original data, and realizing fast and accurate intelligent portrait analysis of various business system capabilities.

[0097] Based on the first embodiment of the present application, in the second embodiment of the present application, the same or similar contents as those in the above-mentioned embodiment 1 can be referred to the above introduction, and will not be repeated in the following. Figure 1 and Figure 2 , Figure 2 A flowchart diagram is provided for Example 2 of the method for constructing a prediction model for the business platform capability portrait of this application.

[0098] In this example, step S30 includes steps S31 to S33:

[0099] Step S31: According to the feature vector extracted by the feature extractor and the labeled data, the Bayesian network is trained using the lower bound of the evidence of variational inference as the minimization loss function to obtain an uncertain model.

[0100] It should be noted that the Evidence Lower Bound (ELBO) is used as a loss function in variational inference. Its principle is to use a simpler distribution (variational distribution) to approximate a complex posterior distribution that is difficult to handle directly.

[0101] By using it as the minimization loss function, the variational distribution can be continuously brought close to the true posterior distribution, thereby achieving effective inference and learning of the Bayesian network and obtaining an uncertain model.

[0102] Specifically, for ease of understanding, the process of training an uncertain model is as follows:

[0103] The reduced-dimensional feature vector extracted by the fine-tuned RoBERTa model can be used as input, and the feature vector can be processed using the Bayesian network. At this time, the weights and biases of the model can be randomly initialized using a normal distribution, and the Bayesian network can be trained using Adam as the optimizer and the Evidence Lower Bound of Variational Inference (ELBO) as the minimization loss function.

[0104] Among them, ELBO consists of two parts: likelihood term (data log likelihood) and regularization term (KL divergence), that is The uncertain model is trained by minimizing the loss function, where θ is the model parameter, is the observed data, i.e., the capability profile of the business middle office system to be predicted, p is the prior probability distribution of the capability profile of the business middle office system, q is the controllable distribution, and its distribution can be dynamically adjusted by the model parameters during the learning process. is a logarithmic function used to calculate the logarithmic likelihood of data, and KL is the KL divergence calculation function.

[0105] Step S32: inferring the data to be labeled of the system operation data through the uncertain model, and outputting labels and uncertainties.

[0106] It should be noted that the data to be labeled is the unlabeled data in the system operation data that has not been preprocessed in step S10. These data do not have clear classifications and labels, and can be inferred by an uncertain model to learn and train the model.

[0107] It should be noted that the label is the image label that the uncertain model automatically assigns to the input data to be labeled after training and learning. Uncertainty is the uncertain probability predicted by the uncertain model for this image label.

[0108] Step S33: According to the label and the uncertainty, the uncertain model is iteratively trained to obtain a prediction model of the business middle station capability portrait.

[0109] It should be noted that after obtaining the label and uncertainty, the output data with low uncertainty can be adopted as the label of the data to be labeled, and the data with high uncertainty can be manually confirmed, and then iterative training can be performed. It is also possible to set an uncertainty threshold and iteratively train the data within the uncertainty threshold. Other methods are also possible, and this embodiment does not limit this.

[0110] In this implementation, a classification model with uncertainty prediction, namely a Bayesian network, is trained based on the existing labeled data and extracted feature vectors. The Bayesian network is used to perform reasoning on the data to be labeled, and labels and uncertainties are output. Outputs with low uncertainty are adopted as labels for the data to be labeled, and data with high uncertainty are manually confirmed and iterative training is performed. For the iterative updates of the business middle-office system, the above-mentioned autonomous learning combined with uncertain learning is used to update the model. Compared with the business middle-office capability portrait construction solution based on the traditional machine learning model and the business middle-office capability portrait construction solution based on natural language processing, the computing power cost and time cost of updating the business middle-office capability portrait are lower.

[0111] In a feasible implementation, the step S33 includes the steps of: manually labeling the data to be labeled according to the uncertainty to obtain manually labeled data; training the uncertain model based on the label and the manually labeled data to obtain an output result; when the output result does not reach a preset stop condition, returning to execute the step of training the Bayesian network based on the feature vector extracted by the feature extractor and the labeled data, using the lower bound of the evidence of variational inference as the minimization loss function; until the output result reaches the preset stop condition, the model corresponding to the preset stop condition is used as the prediction model of the business middle station capability portrait.

[0112] It should be noted that the preset stop condition is the condition for stopping model training, which can be reaching a preset training batch or reaching a performance indicator, and this embodiment does not limit this.

[0113] Specifically, for ease of understanding, after outputting the label and uncertainty, the data to be labeled can be manually labeled according to the uncertainty to obtain manually labeled data. Then the manually labeled data and the above-mentioned label are used as training data for training. After this round of training, return to step S31 and execute the next round of loops, wherein the input of the next round of loops is the manually labeled data and labels of the previous round.

[0114] Among them, before manual labeling, some Dropout layers can be added to the traditional Bayesian network to randomly discard some parameters of the feedforward neural network. Each network will give a set of predicted probability values, and at the same time, a variance calculation is performed on this set of probability values ​​to express the uncertainty of the prediction. When a sample has been expressed with different Dropout parameters, the final probability value is different. The larger the variance of the probability value, the smaller the probability certainty in the learning process. Finally, the data to be labeled with high uncertainty can be manually labeled, and the data with high certainty can be directly adopted by the model labeling results.

[0115] In this implementation, the accuracy of the model can be further improved through multiple rounds of cyclic training.

[0116] In the technical solution provided in this embodiment, a learning framework for autonomous learning and uncertain learning is designed. The feature vector extracted by the fine-tuned RoBERTa model is used as input to train a Bayesian classification model with uncertainty prediction. The output with low uncertainty is adopted as the label of the data to be labeled, and the data with high uncertainty is manually confirmed. Iterative training is performed to achieve low-cost portrait construction and iteration, and fast and accurate intelligent portrait analysis, which can enable business personnel to have a clearer, intuitive and accurate understanding of the existing capabilities of the middle platform, and can also efficiently build application scenarios and solutions.

[0117] Based on the above embodiments of the present application, in the third embodiment of the present application, the same or similar contents as those in the above embodiments can be referred to the above introduction, and will not be described in detail later. Figure 3 , Figure 3 A flowchart diagram is provided for Example 3 of the method for constructing a prediction model for the business platform capability portrait of this application.

[0118] In the steps of this example, when the preset stop condition is a performance indicator, the step of using the model corresponding to the preset stop condition as the prediction model of the business middle station capability portrait until the output result reaches the preset stop condition includes: determining the actual number of business middle station capability portraits in the business middle station system; determining the portrait coverage and accuracy based on the actual number and the predicted number of business middle station capability portraits in the output result; when the portrait coverage and accuracy reach the performance indicator, using the model corresponding to the performance indicator as the prediction model of the business middle station capability portrait.

[0119] It should be noted that the real number is the number of real business middle office capability portraits in the business middle office system. The predicted number is the number of business middle office capability portraits predicted by the trained uncertain model.

[0120] It should be noted that the portrait coverage is the coverage ratio between the number of predicted business middle station capability portraits and the number of actual business middle station capability portraits. The accuracy rate is the correctness of the predicted business middle station capability portraits.

[0121] The portrait coverage and accuracy (i.e., performance indicators) can reflect the performance of the uncertain model after training. When the portrait coverage and accuracy predicted by the uncertain model reach the preset threshold (for example, 90%), the training is stopped and the current model is saved as a prediction model for the business middle-end capability portrait.

[0122] Specifically, we can first estimate the output results of the model at each training stage, that is, the prediction of the uncertain model on the capability portrait of each business middle station. Calculate the portrait coverage and accuracy based on the output results and the manually labeled business middle station capability portrait. Evaluate the model performance based on the portrait coverage and accuracy, and calculate the ratio of the portrait extracted for the specific business middle station capability to the manually labeled business middle station capability portrait. The portrait coverage is calculated as follows:

[0123]

[0124] Among them, TP is the real business middle-office capability portrait and the model predicts the number of this portrait, and FN is the real business middle-office capability portrait but the model does not predict the number of this portrait.

[0125] The accuracy is calculated as follows:

[0126]

[0127] Among them, Predict True The number of portraits predicted by the model and the portraits are real business middle-end capability portraits, Predict False The number of profiles predicted by the model but not the actual business platform capability profiles. If the model performance meets the standard, save the model; if the model performance does not meet the standard, continue training.

[0128] In the technical solution provided in this embodiment, by calculating the coverage and accuracy of the portrait, if the model performance does not meet the standard, continue training, so as to improve the accuracy of the prediction model. By using artificial intelligence for prediction to build the business middle station capability portrait, the rule-based business middle station capability portrait construction scheme and the business middle station capability portrait construction scheme based on statistical analysis are abandoned, reducing the dependence on the knowledge and experience of business experts and the dependence on the quality and integrity of the original data.

[0129] It should be noted that the above examples are only used to understand the present application and do not constitute a limitation on the method for constructing a predictive model for the business platform capability portrait of the present application. More simple transformations based on this technical concept are all within the scope of protection of the present application.

[0130] This application also provides a prediction model construction device for business middle station capability portrait, please refer to Figure 4 , Figure 4 This is a schematic diagram of the module structure of a device for constructing a prediction model for a business middle station capability portrait in an embodiment of the present application; the device for constructing a prediction model for a business middle station capability portrait includes:

[0131] The data preprocessing module 401 is used to preprocess the system operation data of the business middle platform system to obtain the labeled data;

[0132] A feature extraction module 402 is used to perform a pre-training task on the labeled data through a pre-trained language model to obtain a feature extractor, wherein the pre-training task includes log prediction based on system operation logs and prediction of association between logs and business domains;

[0133] The model training module 403 is used to iteratively train the Bayesian network based on the feature vector extracted by the feature extractor and the labeled data to obtain a prediction model for the business middle station capability portrait, and the Bayesian network is a model with uncertainty prediction.

[0134] In this embodiment, the feature extraction module 402 is also used to load the pre-trained language model according to the pre-trained task to obtain a corresponding task prediction model; use the cross entropy loss as the loss function; based on the labeled data, optimize the task prediction model through the loss function to obtain a feature extractor; extract the feature vector of the system operation data according to the feature extractor.

[0135] Furthermore, the model training module 403 is also used to train the Bayesian network based on the feature vector extracted by the feature extractor and the labeled data, using the lower bound of the evidence of variational inference as the minimization loss function to obtain an uncertain model; use the uncertain model to infer the data to be labeled of the system operation data, and output labels and uncertainties; and iteratively train the uncertain model based on the labels and uncertainties to obtain a predictive model for the business middle station capability portrait.

[0136] Furthermore, the model training module 403 is also used to manually label the data to be labeled according to the uncertainty to obtain manually labeled data; train the uncertain model based on the label and the manually labeled data to obtain an output result; when the output result does not reach a preset stopping condition, return to execute the step of training the Bayesian network based on the feature vector extracted by the feature extractor and the labeled data, using the lower bound of the evidence of variational inference as the minimization loss function; until the output result reaches the preset stopping condition, the model corresponding to the preset stopping condition is used as the prediction model of the business middle station capability portrait.

[0137] Furthermore, when the preset stop condition is a performance indicator, the model training module 403 is also used to determine the actual number of business middle-office capability portraits in the business middle-office system; determine the portrait coverage and accuracy based on the actual number and the predicted number of business middle-office capability portraits in the output result; when the portrait coverage and accuracy reach the performance indicator, the model corresponding to the performance indicator will be used as the prediction model of the business middle-office capability portrait.

[0138] Furthermore, the data preprocessing module 401 is also used to obtain system operation data of the business middle-end system, and the system operation data includes user behavior logs, system basic data, system functional data and operation and maintenance data; the user behavior logs, the system basic data, the system functional data and the operation and maintenance data are cleaned to obtain cleaned data; and the cleaned data are labeled with portrait labels to obtain labeled data.

[0139] Other embodiments or specific implementation methods of the device for building a predictive model for the business platform capability portrait of the present application can refer to the above-mentioned method embodiments and will not be repeated here.

[0140] The prediction model construction device for the business middle station capability portrait provided in the present application adopts the prediction model construction method for the business middle station capability portrait in the above-mentioned embodiment, which can solve the technical problem that the existing business middle station capability portrait solution is difficult to adapt to the changes in the business system and there are some business system capabilities that are difficult to cover. Compared with the prior art, the beneficial effects of the prediction model construction device for the business middle station capability portrait provided in the present application are the same as the beneficial effects of the prediction model construction method for the business middle station capability portrait provided in the above-mentioned embodiment, and the other technical features in the prediction model construction device for the business middle station capability portrait are the same as the features disclosed in the above-mentioned embodiment method, which will not be repeated here.

[0141] The present application provides a server or a device specifically used for constructing a prediction model for a business middle-station capability portrait. The prediction model construction device for the business middle-station capability portrait includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the prediction model construction method for the business middle-station capability portrait in the above-mentioned embodiment one.

[0142] Reference below Figure 5 , Figure 5 The device structure diagram of the hardware operating environment involved in the method for constructing a prediction model of the business middle station capability portrait in the embodiment of the present application shows a structural diagram of a device suitable for implementing the prediction model construction of the business middle station capability portrait in the embodiment of the present application. The prediction model construction device for the business middle station capability portrait in the embodiment of the present application may include but is not limited to mobile terminals such as mobile phones, laptops, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Descriptions: tablet computers), PMPs (Portable Media Players: portable multimedia players), vehicle-mounted terminals (such as vehicle-mounted navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc. Figure 5 The predictive model building device for the business middle-end capability portrait shown is merely an example and should not bring any limitations to the functions and scope of use of the embodiments of the present application.

[0143] like Figure 5As shown, the predictive model building device for the business middle station capability portrait may include a processing device 1001 (such as a central processing unit, a graphics processor, etc.), which can perform various appropriate actions and processes according to the program stored in the read-only memory (ROM: Read Only Memory) 1002 or the program loaded from the storage device 1003 to the random access memory (RAM: Random Access Memory) 1004. Various programs and data required for the operation of the predictive model building device for the business middle station capability portrait are also stored in RAM1004. The processing device 1001, ROM1002 and RAM1004 are connected to each other via a bus 1005. The input / output (I / O) interface 1006 is also connected to the bus. Typically, the following systems can be connected to the I / O interface 1006: input devices 1007 including, for example, a touch screen, a touchpad, a keyboard, a mouse, an image sensor, a microphone, an accelerometer, a gyroscope, etc.; output devices 1008 including, for example, a liquid crystal display (LCD: Liquid Crystal Display), a speaker, a vibrator, etc.; storage devices 1003 including, for example, a magnetic tape, a hard disk, etc.; and communication devices 1009. The communication device 1009 can allow the predictive model building device for the business middle station capability portrait to communicate wirelessly or wired with other devices to exchange data. Although the figure shows a predictive model building device for the business middle station capability portrait with various systems, it should be understood that it is not required to implement or have all the systems shown. More or fewer systems may be implemented or provided alternatively.

[0144] In particular, according to the embodiments disclosed in the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the embodiments disclosed in the present application include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network through a communication device, or installed from a storage device 1003, or installed from a ROM 1002. When the computer program is executed by the processing device 1001, the above-mentioned functions defined in the method of the embodiment disclosed in the present application are executed.

[0145] The predictive model construction device for the business middle station capability portrait provided in the present application adopts the predictive model construction method for the business middle station capability portrait in the above-mentioned embodiment, which can solve the technical problem that the existing business middle station capability portrait solution is difficult to adapt to the changes in the business system and there are some business system capabilities that are difficult to cover. Compared with the prior art, the beneficial effects of the predictive model construction device for the business middle station capability portrait provided in the present application are the same as the beneficial effects of the predictive model construction method for the business middle station capability portrait provided in the above-mentioned embodiment, and the other technical features in the predictive model construction device for the business middle station capability portrait are the same as the features disclosed in the method of the previous embodiment, which will not be repeated here.

[0146] It should be understood that the various parts disclosed in this application can be implemented by hardware, software, firmware or a combination thereof. In the description of the above embodiments, specific features, structures, materials or characteristics can be combined in any one or more embodiments or examples in a suitable manner.

[0147] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0148] The present application provides a computer-readable storage medium having computer-readable program instructions (i.e., computer programs) stored thereon, and the computer-readable program instructions are used to execute the method for constructing a predictive model for the business middle-office capability portrait in the above-mentioned embodiment.

[0149] The computer-readable storage medium provided in the present application may be, for example, a USB flash drive, but is not limited to electrical, magnetic, optical, electromagnetic, infrared or semiconductor systems, systems or devices, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, system or device. The program code contained on the computer-readable storage medium may be transmitted using any appropriate medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination of the above.

[0150] The above-mentioned computer-readable storage medium may be included in the prediction model construction device of the business middle platform capability portrait; or it may exist independently without being assembled into the prediction model construction device of the business middle platform capability portrait.

[0151] The above-mentioned computer-readable storage medium carries one or more programs. When the above-mentioned one or more programs are executed by the prediction model building device of the business middle station capability portrait, the prediction model building device of the business middle station capability portrait: pre-processes the system operation data of the business middle station system to obtain labeled data; performs pre-training tasks on the labeled data through a pre-trained language model to obtain a feature extractor, and the pre-training tasks include log prediction based on the system operation log and prediction of the relationship between the log and the business field; iteratively trains the Bayesian network based on the feature vector extracted by the feature extractor and the labeled data to obtain a prediction model of the business middle station capability portrait, and the Bayesian network is a model with uncertainty prediction.

[0152] Computer program code for performing the operations of the present application may be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0153] The flowchart and block diagram in the accompanying drawings illustrate the possible architecture, functions and operations of the system, method and computer program product according to various embodiments of the present application. In this regard, each box in the flowchart or block diagram can represent a module, a program segment or a part of a code, and the module, a program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flowchart and the combination of boxes in the block diagram and / or flowchart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0154] The modules involved in the embodiments described in the present application may be implemented by software or hardware, wherein the name of the module does not constitute a limitation on the unit itself in some cases.

[0155] The readable storage medium provided in this application is a computer-readable storage medium, which stores computer-readable program instructions (i.e., computer programs) for executing the above-mentioned method for building a prediction model for the business middle station capability portrait, which can solve the problem that the existing business middle station capability portrait solution is difficult to adapt to changes in the business system, and there will be some technical problems that are difficult to cover the capabilities of the business system. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided in this application are the same as the beneficial effects of the method for building a prediction model for the business middle station capability portrait provided in the above-mentioned embodiment, which will not be repeated here.

[0156] The present application also provides a computer program product, including a computer program, which, when executed by a processor, implements the steps of the method for constructing a predictive model for the business middle station capability portrait as described above.

[0157] The computer program product provided by this application can solve the problem that the existing business middle station capability portrait solution is difficult to adapt to changes in the business system, and there will be some technical problems that the business system capabilities are difficult to cover. Compared with the prior art, the beneficial effects of the computer program product provided by this application are the same as the beneficial effects of the prediction model construction method for the business middle station capability portrait provided in the above embodiment, which will not be repeated here.

[0158] The above descriptions are only some embodiments of the present application, and are not intended to limit the patent scope of the present application. All equivalent structural changes made using the contents of the present application specification and drawings under the technical concept of the present application, or direct / indirect applications in other related technical fields are included in the patent protection scope of the present application.

Claims

1. A method for constructing a prediction model for a business middle station capability portrait, characterized in that: The method includes: Pre-process the system operation data of the business middle-end system to obtain labeled data; Performing a pre-training task on the labeled data through a pre-trained language model to obtain a feature extractor, wherein the pre-training task includes log prediction based on system operation logs and prediction of association between logs and business fields; Based on the feature vector extracted by the feature extractor and the labeled data, the Bayesian network is iteratively trained to obtain a prediction model for the business middle station capability portrait, and the Bayesian network is a model with uncertainty prediction.

2. The method according to claim 1, characterized in that The step of performing a pre-training task on the labeled data by using a pre-trained language model to obtain a feature extractor comprises: Load the pre-trained language model according to the pre-trained task to obtain the corresponding task prediction model; Use cross entropy loss as the loss function; Based on the labeled data, optimizing the task prediction model through the loss function to obtain a feature extractor; A feature vector of the system operation data is extracted according to the feature extractor.

3. The method according to claim 1, characterized in that The step of iteratively training the Bayesian network based on the feature vector extracted by the feature extractor and the labeled data to obtain a prediction model of the business middle station capability portrait includes: According to the feature vector extracted by the feature extractor and the labeled data, the Bayesian network is trained using the lower bound of the evidence of variational inference as a minimization loss function to obtain an uncertain model; Reasoning on the unlabeled data of the system operation data by using the uncertain model, and outputting labels and uncertainties; According to the label and the uncertainty, the uncertain model is iteratively trained to obtain a prediction model of the business middle station capability portrait.

4. The method according to claim 3, characterized in that The step of iteratively training the uncertain model according to the label and the uncertainty to obtain a prediction model of the business middle station capability portrait includes: Manually labeling the data to be labeled according to the uncertainty to obtain manually labeled data; Training the uncertain model based on the label and the manually annotated data to obtain an output result; When the output result does not meet the preset stop condition, returning to execute the step of training the Bayesian network based on the feature vector extracted by the feature extractor and the labeled data, using the lower bound of the evidence of variational inference as the minimization loss function; Until the output result reaches the preset stop condition, the model corresponding to the preset stop condition will be used as the prediction model of the business middle station capability portrait.

5. The method according to claim 4, characterized in that When the preset stop condition is a performance indicator, the step of using the model corresponding to the preset stop condition as the prediction model of the business middle station capability portrait until the output result reaches the preset stop condition includes: Determine the actual number of business middle office capability portraits in the business middle office system; Determine the coverage and accuracy of the portrait according to the actual number and the predicted number of business middle station capability portraits in the output result; When the portrait coverage and accuracy reach the performance indicators, the model corresponding to the performance indicators will be used as the prediction model of the business middle platform capability portrait.

6. The method according to any one of claims 1 to 5, characterized in that The step of preprocessing the system operation data of the business middle station system to obtain the labeled data includes: Obtaining system operation data of the business middle office system, wherein the system operation data includes user behavior logs, system basic data, system functional data and operation and maintenance data; Cleaning the user behavior log, the system basic data, the system functional data, and the operation and maintenance data to obtain cleaned data; The cleaned data are respectively labeled with image labels to obtain labeled data.

7. A prediction model construction device for business middle station capability portrait, characterized in that: The device comprises: The data preprocessing module is used to preprocess the system operation data of the business middle office system to obtain labeled data; A feature extraction module, used to perform a pre-training task on the labeled data through a pre-trained language model to obtain a feature extractor, wherein the pre-training task includes log prediction based on system operation logs and prediction of association between logs and business domains; The model training module is used to iteratively train the Bayesian network based on the feature vector extracted by the feature extractor and the labeled data to obtain a prediction model for the business middle station capability portrait, and the Bayesian network is a model with uncertainty prediction.

8. A server, characterized in that: The server includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program is configured to implement the steps of a method for constructing a predictive model for a business middle station capability portrait as described in any one of claims 1 to 6.

9. A storage medium, characterized in that: The storage medium is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, the steps of the method for constructing a prediction model of the business middle station capability portrait as described in any one of claims 1 to 6 are implemented.

10. A computer program product, characterized in that The computer program product includes a computer program, which, when executed by a processor, implements the steps of the method for constructing a predictive model for the business middle station capability portrait as described in any one of claims 1 to 6.

Images