Method for training risk prediction model based on cross-domain text and risk prediction method

Abstract

The embodiment of the invention provides a method for training a risk prediction model based on a cross-domain text and a risk prediction method, and the method for training the risk prediction model based on the cross-domain text comprises the steps: reading global text data from a global data source associated with a part change task, reading local text data matched with the global text data from a local data source associated with the part change task; inputting the global text data and the local text data into an initial risk prediction model, and fusing the global text data and the local text data through a fusion module in the initial risk prediction model; processing a fusion result through a prediction module in the initial risk prediction model to obtain prediction risk information output by the initial risk prediction model; and performing parameter adjustment on the initial risk prediction model based on target risk information corresponding to the global text data and the prediction risk information until a risk prediction model meeting a training stop condition is obtained.

Description

Technical Field

[0001] The embodiments of this specification relate to the field of deep learning technology, and in particular to a method for training a risk prediction model based on cross-domain text and a risk prediction method. Background Art

[0002] With the development of computer technology, more and more industries have begun to automate. Automated equipment can replace some manual labor and complete product manufacturing more efficiently and quickly. However, with the automation of some production lines, risk events in production and manufacturing will also occur frequently. In the existing technology, in industrial production and manufacturing scenarios, most risk estimates remain at product quality-related risks, and pay more attention to the overall production line risks brought about by quality risks. The amount of information contained is relatively single and one-sided, and risk judgment basically belongs to the second half of the production line, which cannot be discovered and avoided in time. Moreover, the realization of risk estimation relies more on the strategies given by experts, and a lot of time is needed in the early stage to output strategy standards, resulting in high risk assessment costs. Therefore, an effective solution is urgently needed to solve the above problems. Summary of the invention

[0003] In view of this, an embodiment of this specification provides a method for training a risk prediction model based on cross-domain text. One or more embodiments of this specification also relate to a device for training a risk prediction model based on cross-domain text, a risk prediction method, a risk prediction device, a computing device, a computer-readable storage medium, and a computer program to solve the technical defects existing in the prior art.

[0004] According to a first aspect of an embodiment of this specification, a method for training a risk prediction model based on cross-domain text is provided, comprising:

[0005] Reading global text data from a global data source associated with a part change task, and reading local text data matching the global text data from a local data source associated with the part change task;

[0006] Inputting the global text data and the local text data into an initial risk prediction model, and fusing the global text data and the local text data through a fusion module in the initial risk prediction model;

[0007] Processing the fusion result through the prediction module in the initial risk prediction model to obtain the predicted risk information output by the initial risk prediction model;

[0008] Based on the target risk information and the predicted risk information corresponding to the global text data, the initial risk prediction model is adjusted until a risk prediction model that meets the training stop condition is obtained.

[0009] Optionally, the reading of global text data from a global data source associated with the part change task, and the reading of local text data matching the global text data from a local data source associated with the part change task, include:

[0010] Determine a global data source and a local data source associated with the part change task, wherein the global data source is used to record text data associated with all part change processes in the part change task, and the local data source is used to record text data associated with part change anomalies in all part change processes;

[0011] The global text data is read from the global data source, and the local data source is searched according to the global text data to obtain the local text data matching the global text data.

[0012] Optionally, searching the local data source according to the global text data to obtain the local text data matching the global text data includes:

[0013] Searching the local data source according to the global text data, and detecting whether there is text data matching the global text data in the local data source according to the search result;

[0014] If so, reading text data matching the global text data from the local data source as the local text data;

[0015] If not, reference text data is read from the local data source as the local text data.

[0016] Optionally, fusing the global text data and the local text data through a fusion module in the initial risk prediction model includes:

[0017] The global text data and the local text data are respectively encoded by the encoding unit included in the fusion module in the initial risk prediction model to obtain a global text vector corresponding to the global text data and a local text vector corresponding to the local text data;

[0018] The global text vector and the local text vector are fused by a fusion unit included in the fusion module to obtain a text fusion vector;

[0019] The step of processing the fusion result by the prediction module in the initial risk prediction model to obtain the predicted risk information output by the initial risk prediction model includes:

[0020] The text fusion vector is processed by the prediction module in the initial risk prediction model to obtain the predicted risk information output by the initial risk prediction model.

[0021] Optionally, the determination of the global text vector includes:

[0022] In the case where the global text data is a discrete data type, encoding the global text data by a first sub-encoding unit in the encoding unit to obtain the global text vector;

[0023] In the case where the global text data is a continuous data type, the global text data is converted into target global text data of a corresponding discrete data type, and the target global text data is encoded by a second sub-encoding unit in the encoding unit to obtain the global text vector.

[0024] Optionally, the determination of the local text vector includes:

[0025] By using the word segmentation unit in the encoding unit, the local text data is segmented according to a pre-generated word list to obtain a plurality of word units;

[0026] The plurality of word units are encoded by a third sub-encoding unit in the encoding unit to obtain the local text vector.

[0027] Optionally, adjusting the parameters of the initial risk prediction model based on the target risk information and the predicted risk information corresponding to the global text data until a risk prediction model that meets the training stop condition is obtained includes:

[0028] Detecting the global text data according to the business logic rules corresponding to the part change task, and determining the target risk information corresponding to the global text data according to the detection result;

[0029] A loss value is calculated based on the target risk information and the predicted risk information, and the initial risk prediction model is adjusted according to the loss value until a risk prediction model that meets the training stop condition is obtained.

[0030] According to a second aspect of the embodiments of this specification, a risk prediction method is provided, comprising:

[0031] Obtain part description data associated with the target part change task;

[0032] Inputting the part description data into the risk prediction model in the above method for processing to obtain part risk information;

[0033] The prompt information corresponding to the target part change task is updated according to the part risk information, and the updated prompt information is displayed.

[0034] According to a third aspect of an embodiment of this specification, there is provided a device for training a risk prediction model based on cross-domain text, comprising:

[0035] A reading module, configured to read global text data from a global data source associated with a part change task, and to read local text data matching the global text data from a local data source associated with the part change task;

[0036] An input module, configured to input the global text data and the local text data into an initial risk prediction model, and fuse the global text data and the local text data through a fusion module in the initial risk prediction model;

[0037] An output module is configured to process the fusion result through the prediction module in the initial risk prediction model to obtain the predicted risk information output by the initial risk prediction model;

[0038] The parameter adjustment module is configured to adjust the parameters of the initial risk prediction model based on the target risk information and the predicted risk information corresponding to the global text data until a risk prediction model that meets the training stop condition is obtained.

[0039] According to a fourth aspect of the embodiments of this specification, there is provided a risk prediction device, comprising:

[0040] A data acquisition module is configured to acquire part description data associated with a target part change task;

[0041] A model prediction module is configured to input the part description data into the risk prediction model in the above method for processing to obtain part risk information;

[0042] The update information module is configured to update the prompt information corresponding to the target part change task according to the part risk information, and display the updated prompt information.

[0043] According to a fifth aspect of an embodiment of this specification, a computing device is provided, including:

[0044] Memory and processor;

[0045] The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, the steps of the above-mentioned risk prediction model or risk prediction method based on cross-domain text training are implemented.

[0046] According to the sixth aspect of the embodiments of this specification, a computer-readable storage medium is provided, which stores computer-executable instructions, which, when executed by a processor, implement the steps of the above-mentioned risk prediction model or risk prediction method based on cross-domain text training.

[0047] According to the seventh aspect of the embodiments of this specification, a computer program is provided, wherein when the computer program is executed in a computer, the computer is caused to execute the steps of the above-mentioned risk prediction model based on cross-domain text training or risk prediction method.

[0048] The method for training a risk prediction model based on cross-domain text provided in this embodiment is to train a model that can replace manual risk assessment for a part change task, so as to improve the real-time and accuracy of risk assessment. The global text data can be read from the global data source associated with the part change task first, and the local text data matched by the global text data can be read from the local data source associated with the part change task; the richness and universality of the samples can be improved by using two parts of data from different data sources and aligned data as samples for training. Thereafter, the global text data and the local text data are input into the initial risk prediction model, and the global text data and the local text data are fused through the fusion module in the initial risk prediction model; and the fusion result is processed by the prediction module in the initial risk prediction model, and the predicted risk information output by the initial risk prediction model can be obtained according to the processing result; on this basis, the initial risk prediction model can be adjusted based on the target risk information and predicted risk information corresponding to the global text data, until a risk prediction model that meets the training stop condition is obtained. In the model training stage, the cross-domain text can be selected to train the model in the current scenario, so that the model prediction accuracy can be effectively improved. This enables real-time and efficient risk prediction in part change scenarios, thereby effectively reducing dependence on experts. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 It is a schematic diagram of a method for training a risk prediction model based on cross-domain texts provided by an embodiment of this specification;

[0050] Figure 2a is a flowchart of a method for training a risk prediction model based on cross-domain texts provided by an embodiment of this specification;

[0051] Figure 2b It is a schematic diagram of a risk prediction model structure provided by an embodiment of this specification;

[0052] Figure 3 is a flow chart of a risk prediction method provided by an embodiment of this specification;

[0053] Figure 4 It is a structural schematic diagram of a device for training a risk prediction model based on cross-domain texts provided by an embodiment of this specification;

[0054] Figure 5 It is a schematic diagram of the structure of a risk prediction device provided by an embodiment of this specification;

[0055] Figure 6 It is a structural block diagram of a computing device provided by an embodiment of this specification. DETAILED DESCRIPTION

[0056] Many specific details are described in the following description to facilitate a full understanding of this specification. However, this specification can be implemented in many other ways than those described herein, and those skilled in the art can make similar generalizations without violating the connotation of this specification, so this specification is not limited to the specific implementation disclosed below.

[0057] The terms used in one or more embodiments of this specification are only for the purpose of describing specific embodiments, and are not intended to limit one or more embodiments of this specification. The singular forms of "a", "said" and "the" used in one or more embodiments of this specification and the appended claims are also intended to include plural forms, unless the context clearly indicates other meanings. It should also be understood that the term "and / or" used in one or more embodiments of this specification refers to and includes any or all possible combinations of one or more associated listed items.

[0058] It should be understood that although the terms first, second, etc. may be used to describe various information in one or more embodiments of this specification, this information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other. For example, without departing from the scope of one or more embodiments of this specification, the first may also be referred to as the second, and similarly, the second may also be referred to as the first. Depending on the context, the word "if" as used herein may be interpreted as "at the time of" or "when" or "in response to determining".

[0059] In addition, it should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in one or more embodiments of this specification are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with the relevant laws, regulations and standards of the relevant countries and regions, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0060] In this specification, a method for training a risk prediction model based on cross-domain text is provided. This specification also relates to a device for training a risk prediction model based on cross-domain text, a risk prediction method, a risk prediction device, a computing device, and a computer-readable storage medium, which are described in detail one by one in the following embodiments.

[0061] In actual applications, in the production and manufacturing of products, especially in the task of parts change, the risks exist from many aspects, including but not limited to the production cycle, logistics cycle, process allocation, etc. It is impossible to make perfect strategies manually, and different aspects interact with each other, which has a great impact on risk judgment. Therefore, an effective solution is urgently needed to solve the above problems.

[0062] See also Figure 1 As shown in the schematic diagram, the method for training a risk prediction model based on cross-domain text provided in this embodiment is to be able to train a model that can replace manual risk assessment for a part change task, so as to improve the real-time and accuracy of risk assessment. The global text data can be read from the global data source associated with the part change task first, and the local text data matched by the global text data can be read from the local data source associated with the part change task; the richness and universality of the samples can be improved by using two parts of data from different data sources and aligned data as samples for training. Thereafter, the global text data and the local text data are input into the initial risk prediction model, and the global text data and the local text data are fused through the fusion module in the initial risk prediction model; and the fusion result is processed by the prediction module in the initial risk prediction model, and the predicted risk information output by the initial risk prediction model can be obtained according to the processing result; on this basis, the initial risk prediction model can be adjusted based on the target risk information and predicted risk information corresponding to the global text data, until a risk prediction model that meets the training stop condition is obtained. In the model training stage, the cross-domain text can be selected to train the model in the current scene, so that the model prediction accuracy can be effectively improved. This enables real-time and efficient risk prediction in part change scenarios, thereby effectively reducing dependence on experts.

[0063] See also Figure 2a , Figure 2a A flowchart of a method for training a risk prediction model based on cross-domain texts provided according to an embodiment of the present specification is shown, which specifically includes the following steps.

[0064] Step S202: reading global text data from a global data source associated with a part change task, and reading local text data matching the global text data from a local data source associated with the part change task.

[0065] The method based on the cross-domain text training risk prediction model provided in this embodiment can be applied to the risk prediction associated with the parts change tasks of any product production line; for example, the parts change risk prediction of a vehicle production line; another example is the parts change risk prediction of an electrical production line; another example is the parts change risk prediction of an intelligent device production line; this embodiment takes the parts change risk prediction of a vehicle production line as an example to illustrate the method based on the cross-domain text training risk prediction model. For other scenarios that are the same or corresponding, please refer to the same or corresponding description content of this embodiment, and this embodiment will not be elaborated in detail here.

[0066] Specifically, a part change task refers to a task of changing parts for a product in a product manufacturing production line, and the change of product parts will bring many risks, such as production cycle, logistics cycle, process allocation and other risks, resulting in the product being unable to be produced or manufactured normally. Correspondingly, the global data source specifically refers to a data source that records data items corresponding to all changed parts related to the product. The recorded data includes but is not limited to part model information, related vehicle manufacturer information, related vehicle model information, process attribution information, time information, etc., which is used to record all part-related information for use in the model training phase. Correspondingly, global text data is text data generated after reading the data corresponding to a certain part from the global data source.

[0067] Furthermore, the local data source specifically refers to a data source that records data items corresponding to problematic parts among all parts that have changed related to the product, including but not limited to problematic part model information, problem information, abnormal time, etc. It is used to record relevant information about problematic parts. Correspondingly, local text data is text data generated after reading data corresponding to a certain part from a local data source. It should be noted that the data in the global data source and the data in the local data source have different text representations, that is, there is a difference in the text representation of the same data in the two. For example, the text representation of the data corresponding to part 1 is A in the global data source and B in the local data source, where A is not equal to B.

[0068] Based on this, in order to be able to train accurate risk prediction of the changed parts involved in the part change task and improve the real-time performance of risk prediction, the global text data can be read from the global data source associated with the part change task, and the local text data matched by the global text data can be read from the local data source associated with the part change task, so as to realize the reading of the global text data and local text data associated with the same part from two different data sources, and because the two parts of data come from two different data sources, there is a certain difference in text representation between the two. Using this composition sample to train the risk prediction model can realize cross-domain text analysis while integrating multi-dimensional information to complete model training, thereby effectively improving the prediction accuracy of part change risks.

[0069] In an optional embodiment of the present specification, the reading of global text data from a global data source associated with the part change task, and the reading of local text data matching the global text data from a local data source associated with the part change task, include:

[0070] Determine a global data source and a local data source associated with the part change task, wherein the global data source is used to record text data associated with all part change processes in the part change task, and the local data source is used to record text data associated with part change anomalies in all part change processes;

[0071] The global text data is read from the global data source, and the local data source is searched according to the global text data to obtain the local text data matching the global text data.

[0072] Specifically, the text data associated with all parts change processing is the text representation of the data items corresponding to all changed parts associated with the product in the parts change task; correspondingly, the text data associated with part change exceptions is the text representation of the data items corresponding to the parts with problems among all changed parts associated with the product in the parts change task.

[0073] Based on this, in the sample acquisition stage, the global data source and local data source associated with the part change task can be determined first, and the global data source is used to record the text data associated with all part change processes in the part change task, and the local data source is used to record the text data associated with part change exceptions in all part change processes; based on this, the global text data corresponding to any part can be read in the global data source first, and then the local data source can be searched according to the global text data to determine the local text data that matches the global text data based on the search results, so as to select the text representation composition samples of two different data sources for subsequent model training.

[0074] In an optional embodiment of the present specification, searching the local data source according to the global text data to obtain the local text data matching the global text data includes:

[0075] The local data source is searched according to the global text data, and it is detected whether there is text data matching the global text data in the local data source based on the search result; if so, the text data matching the global text data is read from the local data source as the local text data; if not, the reference text data is read from the local data source as the local text data.

[0076] Specifically, the reference text data refers to the text data set for parts that have text data in the global data source but no text data in the local data source. The text data can be a text representation of text descriptions such as "no problem" or "none". In other words, if the part corresponding to the global text data selected in the global data source does not have a part change exception, then the corresponding text data will not exist in the local data source. Therefore, the reference text data can be selected as the local text data for model training.

[0077] Based on this, when the global text data corresponding to any changed part is read in the global data source, the local data source will be searched according to the global text data, and the local data source will be detected based on the search results to see whether there is text data matching the global text data; that is, to detect whether there is text data of the changed part corresponding to the global text data in the local data source; if so, it means that there is a part change anomaly in the changed part, and the text data of the part is recorded in the local data source, then the text data matching the global text data can be read in the local data source as the local text data. If not, it means that there is no anomaly in the changed part, and the text data of the part is not recorded in the local data source. In order to match the global text data for subsequent model training and avoid model overfitting, the benchmark text data can be read in the local data source, that is, the preset text data can be selected as the local text data to facilitate subsequent model training.

[0078] For example, in the scenario of vehicle parts change tasks, in order to predict the risks after vehicle parts are changed, the risk prediction model can be trained in combination with the data associated with the vehicle parts change task. First, determine the first data source B1 that records all parts change information in the vehicle parts change task, and the second data source B2 that records the information of abnormal parts changes. Secondly, extract the text data B1-T corresponding to part 1 from the first data source B1, and detect whether there is text data corresponding to part 1 in the second data source B2. If it exists, it means that there is an abnormality in the change of part 1, and the text data B2-T corresponding to part 1 can be read from the second data source B2. If it does not exist, it means that the change of part 1 is normal, and "no problem" can be selected as the text data B2-T corresponding to part 1. It is used to train the risk prediction model after aligning the text data B1-T and the text data B2-T.

[0079] In summary, by extracting global text data and local text data from two different data sources to form samples, it is possible to train the model with richer text data to improve the model's prediction ability. At the same time, setting benchmark text data for parts that do not have local text data can ensure that any sample has aligned local text data to avoid overfitting of the trained model.

[0080] Step S204: input the global text data and the local text data into an initial risk prediction model, and fuse the global text data and the local text data through a fusion module in the initial risk prediction model.

[0081] Specifically, after obtaining the global text data and local text data as mentioned above, the model can be further trained using the acquired samples, that is, the global text data and the local text data are input together into the initial risk prediction model to be trained. Since the global text data and the local text data come from two different data sources, there are certain differences in their text representations. Therefore, the global text data and the local text can be fused through the fusion module in the initial risk prediction model first, so that the text data from two different data sources are fused and then used for subsequent risk prediction processing, thereby ensuring that the model can learn risk prediction capabilities in different dimensions to improve the risk prediction accuracy of the model.

[0082] Among them, the initial risk prediction model specifically refers to a model used to predict the risk of part changes. The model includes a fusion module and a prediction module. The fusion module is used to fuse global text data and local text data, and the prediction module is used to process the fusion results and output the prediction results.

[0083] In an optional embodiment of the present specification, the fusing of the global text data and the local text data by a fusion module in the initial risk prediction model includes:

[0084] The global text data and the local text data are encoded respectively through the encoding unit included in the fusion module in the initial risk prediction model to obtain a global text vector corresponding to the global text data and a local text vector corresponding to the local text data; the global text vector and the local text vector are fused through the fusion unit included in the fusion module to obtain a text fusion vector.

[0085] Specifically, the encoding unit refers to a unit that encodes global text data and local text data, and is used to convert global text data into global text vectors, and convert local text data into local text vectors; wherein the global text vector is the encoding vector obtained after encoding the global text data; the local text vector is the encoding vector obtained after encoding the local text data. Correspondingly, the text fusion vector is the result of vector fusion of the local text vector and the global text vector.

[0086] Based on this, after the global text data and the local text data are input into the initial risk prediction model, considering that the global text data and the local text data come from two different data sources, the global text data and the local text data can be encoded separately through the encoding unit contained in the fusion module in the initial risk prediction model, so as to obtain the global text vector corresponding to the global text data and the local text vector corresponding to the local text data according to the encoding result; then, the global text vector and the local text vector are fused using the fusion unit contained in the fusion module, so that the two data from different data sources can be fused together to obtain a text fusion vector, which is convenient for subsequent input into the prediction module for processing.

[0087] In an optional embodiment of the present specification, the determination of the global text vector includes:

[0088] In the case where the global text data is a discrete data type, the global text data is encoded by the first sub-encoding unit in the encoding unit to obtain the global text vector; in the case where the global text data is a continuous data type, the global text data is converted into target global text data corresponding to the discrete data type, and the target global text data is encoded by the second sub-encoding unit in the encoding unit to obtain the global text vector.

[0089] Specifically, the discrete data type specifically refers to the type of global text data as discrete data items, such as global text data containing part numbers, car attributes, etc.; correspondingly, the sub-coding unit specifically refers to the coding unit that encodes the global text data of the discrete data type; correspondingly, the continuous data type specifically refers to the type of global text data as continuous data items, such as global text data containing time nodes, time intervals, etc. Correspondingly, the target global text data specifically refers to the result obtained by converting the global text data of the continuous data type into text data of the discrete data type.

[0090] Based on this, when the global text data is a discrete data type, it means that the global text data obtained at this time can be directly encoded, so the global text data can be encoded by the first sub-encoding unit in the encoding unit to obtain a global text vector; when the global text data is a continuous data type, it means that the data type cannot be directly encoded, then the global text data can be converted into target global text data corresponding to the discrete data type, and then the target global text data can be encoded by the second sub-encoding unit in the encoding unit to obtain a global text vector.

[0091] In practical applications, when converting global text data of continuous data type into target global text data of corresponding discrete data type, the global text data can be discretized based on business settings to obtain target global text data that can be encoded and processed for use. In addition, if the global text data contains two types of data at the same time, that is, data of both discrete data type and continuous data type, the data corresponding to the two types can be encoded separately and then fused to obtain a global text vector corresponding to the global text data.

[0092] In an optional embodiment of the present specification, the determination of the local text vector includes:

[0093] The local text data is segmented according to a pre-generated word list by the word segmentation unit in the encoding unit to obtain a plurality of word units; the plurality of word units are encoded by a third sub-encoding unit in the encoding unit to obtain the local text vector.

[0094] Specifically, the word segmentation unit specifically refers to a unit in the encoding unit for performing word segmentation processing on local text data; accordingly, the word list specifically refers to a word list composed of high-frequency phrases extracted by the initial risk prediction model using a deep learning algorithm, and the word units in the word list are all associated with part change tasks. Accordingly, multiple word units specifically refer to word units obtained after word segmentation processing on local text data. Accordingly, the third sub-encoding unit specifically refers to an encoding unit that encodes multiple word units.

[0095] Based on this, when encoding local text data, the local text data can be first segmented through the word segmentation unit in the encoding unit according to a pre-generated vocabulary, so as to obtain multiple word units based on the word segmentation results; thereafter, the multiple word units are encoded through the third sub-encoding unit in the encoding unit to obtain the local text vector corresponding to the local text data.

[0096] Continuing with the above example, after obtaining the text data B1-T and text data B2-T corresponding to part 1, the text data B1-T and text data B2-T can be input into the risk prediction model. If the text data B1-T is a discrete data item, such as the number of part 1 and the attributes of the vehicle to which it belongs, the one-hot encoding method in the risk prediction model can be used to encode the text data B1-T. If the text data B1-T is a continuous data item, such as the change time and storage time of part 1, it can be discretized based on the business first, and then the discretized result can be encoded using the one-hot encoding method in the risk prediction model. If the text data B1-T contains two types of data at the same time, the two types of data can be encoded separately and then concatenated to obtain the encoding vector VC1 corresponding to the text data B1-T. At the same time, the pre-generated vocabulary in the risk prediction model is used to segment the text data B2-T, and the text is encoded using LSTM (long short-term memory) based on the token of the segmentation, so that the encoding vector VC2 corresponding to the text data B2-T can be obtained. By fusing the encoding vector VC1 and the encoding vector VC2 to obtain the fused vector VC, and then performing subsequent processing, the predicted risk information corresponding to part 1 can be obtained.

[0097] In summary, using different coding units for encoding global text data and local text data respectively can ensure that the encoding results can fully characterize the representation of the corresponding data, and on this basis, the two encoding vectors are fused to achieve cross-domain text fusion, and model training is carried out on this basis, which can enable the model to learn richer risk prediction knowledge, thereby improving the model's prediction ability.

[0098] Step S206, processing the fusion result through the prediction module in the initial risk prediction model to obtain the predicted risk information output by the initial risk prediction model.

[0099] Specifically, after the fusion unit in the initial risk prediction model fuses the global text data and the local text data, further, after the fusion of the two data is completed across domains, the prediction module in the initial risk prediction model can be used to process the fusion result, so that the predicted risk information can be output according to the prediction result, and then the model can be adjusted in combination with the sample label to train a risk prediction model that meets the use requirements. Among them, the predicted risk information specifically refers to the information obtained after the risk prediction of the changed parts corresponding to the global text data, which can be used to characterize the risk probability of the changed parts, and / or the risk type, and / or the risk cause, etc.

[0100] That is to say, the fusion result is processed by the prediction module in the initial risk prediction model to obtain the predicted risk information output by the initial risk prediction model. Specifically, it means that the text fusion vector is processed by the prediction module in the initial risk prediction model to obtain the predicted risk information output by the initial risk prediction model.

[0101] In practical applications, the initial risk prediction model can add a fully connected neural network after the fusion module, set two fully connected layers, the hidden_size of the first layer is equal to 64, and the hidden_size of the second layer is equal to 16; and the activation functions of both layers are relu, and the final model output can be passed through softmax to obtain the risk probability value of the sample, that is, the predicted risk information of the changed parts.

[0102] Following the above example, after obtaining the fusion vector VC, the fusion vector can be input into the prediction module in the risk prediction model for processing, as shown in FIG. Figure 2b As shown in the schematic diagram of the model structure, after the model inputs the data of part 1 from two data sources, it will be encoded through two branches respectively, and after obtaining the encoding vectors corresponding to the two data respectively, they will be fused and predicted, so that the risk probability value S1 corresponding to part 1 can be obtained according to the prediction result and the model can be output, which is convenient for subsequent adjustment of the model parameters in combination with sample labels.

[0103] Step S208, adjusting the parameters of the initial risk prediction model based on the target risk information and the predicted risk information corresponding to the global text data until a risk prediction model that meets the training stop condition is obtained.

[0104] Specifically, after obtaining the predicted risk information output by the model as mentioned above, we can further load the sample labels corresponding to the global text data, that is, the target risk information; on this basis, adjust the parameters of the initial risk prediction model in combination with the target risk information and the predicted risk information, and enter a new round of model training until a risk prediction model that meets the training stop conditions is obtained.

[0105] Among them, the target risk information specifically refers to the sample label generated in combination with the business logic. The sample is the data item of the changed part in each dimension. The target risk information is the information whether the changed part is risky. For example, the change type of the changed part contained in the global text data is A, and the record variables are: change start time, interval time, and arrival time. If the time difference between the change start time and the arrival time is greater than the time difference between the interval time and the arrival time, the label is risky, otherwise it is risk-free. For another example, the change type of the changed part contained in the global text data is B. If the time difference between the interval time and the arrival time is greater than the set threshold, the label is risky, otherwise it is risk-free. Correspondingly, the training stop condition specifically refers to the condition for stopping the training of the risk prediction model, which includes but is not limited to the loss value comparison condition, the number of iterations condition, or the model verification condition. Among them, the loss value comparison condition is to compare the loss value of each round of training of the model with the loss value threshold; the number of iterations condition is to detect whether the number of training times of the model meets the preset training times threshold; the model verification condition is to verify the prediction accuracy of the model using the verification set.

[0106] In an optional embodiment of the present specification, the initial risk prediction model is adjusted based on the target risk information and the predicted risk information corresponding to the global text data until a risk prediction model that meets the training stop condition is obtained, including:

[0107] The global text data is detected according to the business logic rules corresponding to the part change task, and the target risk information corresponding to the global text data is determined according to the detection result; the loss value is calculated based on the target risk information and the predicted risk information, and the initial risk prediction model is adjusted according to the loss value until a risk prediction model that meets the training stop conditions is obtained.

[0108] Specifically, the business logic rule specifically refers to the rule set according to the part change task, which is used to detect the global text data corresponding to the changed part, so as to determine whether the part has a risk based on the detection result, and determine the label corresponding to the global text data based on whether there is a risk, that is, the target risk information. For example, the change type of the changed part is A, and the record variables are: change start time, interval time, and arrival time. It is detected whether the time difference between the change start time and the arrival time is greater than the time difference between the interval time and the arrival time. It is determined whether there is a risk based on the detection result, so as to determine the label of risk and no risk; for another example, the change type of the changed part is B. It is detected whether the time difference between the interval time and the arrival time is greater than the set threshold. It is determined whether there is a risk based on the detection result, so as to determine the label of risk and no risk. In practical applications, business logic rules can be set according to actual needs, and this embodiment does not make any restrictions here.

[0109] Based on this, when determining sample labels, in order to save human resources and improve labeling efficiency, the global text data can be detected according to the business logic rules corresponding to the part change task, and the target risk information corresponding to the global text data can be determined according to the detection results; on this basis, the loss value is calculated based on the target risk information and the predicted risk information, and the initial risk prediction model is adjusted according to the loss value. At this time, it can be detected whether the adjusted model meets the training stop conditions; if not, new samples are selected to continue training. If so, the model at the current stage can be used as a risk prediction model that meets the training stop conditions for use in the application stage.

[0110] Continuing with the above example, after obtaining the risk probability value S1 output by the model, you can combine the pre-set business logic to perform risk detection on the changed parts to detect whether the time difference between the change start time and the arrival time of part 1 is greater than the time difference between the interval time and the arrival time. If it is greater, it is determined that part 1 is at risk; if it is less, it is determined that part 1 is not at risk. Use this as a label and the risk probability value S1 output by the model to adjust the parameters of the risk prediction model, and select new samples to continue training the model after the parameters are adjusted. After n rounds of training, you can get a risk prediction model that can predict the risk of part changes. When any part changes, its corresponding data can be input into the model for prediction to obtain the change risk of the part in real time, avoiding impacts on the production cycle, logistics cycle, and process allocation.

[0111] The method for training a risk prediction model based on cross-domain text provided in this embodiment is to train a model that can replace manual risk assessment for a part change task, so as to improve the real-time and accuracy of risk assessment. The global text data can be read from the global data source associated with the part change task first, and the local text data matched by the global text data can be read from the local data source associated with the part change task; the richness and universality of the samples can be improved by using two parts of data from different data sources and aligned data as samples for training. Thereafter, the global text data and the local text data are input into the initial risk prediction model, and the global text data and the local text data are fused through the fusion module in the initial risk prediction model; and the fusion result is processed by the prediction module in the initial risk prediction model, and the predicted risk information output by the initial risk prediction model can be obtained according to the processing result; on this basis, the initial risk prediction model can be adjusted based on the target risk information and predicted risk information corresponding to the global text data, until a risk prediction model that meets the training stop condition is obtained. In the model training stage, the cross-domain text can be selected to train the model in the current scenario, so that the model prediction accuracy can be effectively improved. This enables real-time and efficient risk prediction in part change scenarios, thereby effectively reducing dependence on experts.

[0112] Corresponding to the above method embodiment, this specification also provides a risk prediction method embodiment, Figure 3 FIG. 1 is a flow chart showing a risk prediction method provided by an embodiment of the present specification. Figure 3 As shown, the method includes:

[0113] Step S302, obtaining part description data associated with the target part change task;

[0114] Step S304, inputting the part description data into the risk prediction model in the above method for processing to obtain part risk information;

[0115] Step S306: updating the prompt information corresponding to the target part change task according to the part risk information, and displaying the updated prompt information.

[0116] The risk prediction method provided in this embodiment corresponds to the above-mentioned method based on cross-domain text training risk prediction model. The same or corresponding description contents can be found in the above-mentioned embodiments, and this embodiment will not be elaborated in detail here.

[0117] Based on this, the part description data specifically refers to the global text data read from the global data source associated with the target part change task, and the global text data is all data items involved in the changed part to be detected. Correspondingly, the part risk information specifically refers to the risk value of the changed part to be detected. Correspondingly, the prompt information specifically refers to the information that prompts whether the changed part to be detected has risks, such as indications of different colors, etc. A red light indicates that the part has risks, and a green light indicates that the part has no risks. The prompt information can also be displayed in other forms, and this embodiment does not make any limitation here.

[0118] Based on this, after obtaining the part description data associated with the target part change task, the part description data can be input into the risk prediction model in the above method for processing, so as to predict the part risk information corresponding to the part according to the trained model; thereafter, in order to be able to remind the staff in time, the prompt information corresponding to the target part change task can be updated according to the part risk information, and the updated prompt information can be displayed, so that the staff can pay attention to the risk status of any changed parts in time.

[0119] In summary, by using risk prediction models instead of manually configuring rules to complete risk prediction of changed parts, the efficiency and accuracy of risk prediction can be effectively improved, and the consumption of human resources can be reduced to ensure that risk detection can be completed quickly in part change tasks and reduce the impact of risk events.

[0120] Corresponding to the above method embodiment, this specification also provides an apparatus embodiment for training a risk prediction model based on cross-domain text, Figure 4 FIG. 1 is a schematic diagram showing a structure of a device for training a risk prediction model based on cross-domain text according to an embodiment of the present specification. Figure 4 As shown, the device comprises:

[0121] A reading module 402 is configured to read global text data from a global data source associated with a part change task, and read local text data matching the global text data from a local data source associated with the part change task;

[0122] An input module 404 is configured to input the global text data and the local text data into an initial risk prediction model, and fuse the global text data and the local text data through a fusion module in the initial risk prediction model;

[0123] The output module 406 is configured to process the fusion result through the prediction module in the initial risk prediction model to obtain the predicted risk information output by the initial risk prediction model;

[0124] The parameter adjustment module 408 is configured to adjust the parameters of the initial risk prediction model based on the target risk information and the predicted risk information corresponding to the global text data until a risk prediction model that meets the training stop condition is obtained.

[0125] In an optional embodiment, the reading module 402 is further configured to:

[0126] Determine a global data source and a local data source associated with the part change task, wherein the global data source is used to record text data associated with all part change processes in the part change task, and the local data source is used to record text data associated with part change exceptions in all part change processes; read global text data from the global data source, and search the local data source according to the global text data to obtain local text data matching the global text data.

[0127] In an optional embodiment, the reading module 402 is further configured to:

[0128] The local data source is searched according to the global text data, and it is detected whether there is text data matching the global text data in the local data source based on the search result; if so, the text data matching the global text data is read from the local data source as the local text data; if not, the reference text data is read from the local data source as the local text data.

[0129] In an optional embodiment, the input module 404 is further configured to:

[0130] The global text data and the local text data are encoded respectively by the encoding unit included in the fusion module in the initial risk prediction model to obtain a global text vector corresponding to the global text data and a local text vector corresponding to the local text data; the global text vector and the local text vector are fused by the fusion unit included in the fusion module to obtain a text fusion vector; the fusion result is processed by the prediction module in the initial risk prediction model to obtain the predicted risk information output by the initial risk prediction model, including: the text fusion vector is processed by the prediction module in the initial risk prediction model to obtain the predicted risk information output by the initial risk prediction model.

[0131] In an optional embodiment, the input module 404 is further configured to:

[0132] In the case where the global text data is a discrete data type, the global text data is encoded by the first sub-encoding unit in the encoding unit to obtain the global text vector; in the case where the global text data is a continuous data type, the global text data is converted into target global text data corresponding to the discrete data type, and the target global text data is encoded by the second sub-encoding unit in the encoding unit to obtain the global text vector.

[0133] In an optional embodiment, the input module 404 is further configured to:

[0134] The local text data is segmented according to a pre-generated word list by the word segmentation unit in the encoding unit to obtain a plurality of word units; the plurality of word units are encoded by a third sub-encoding unit in the encoding unit to obtain the local text vector.

[0135] In an optional embodiment, the parameter adjustment module 408 is further configured to:

[0136] The global text data is detected according to the business logic rules corresponding to the part change task, and the target risk information corresponding to the global text data is determined according to the detection result; the loss value is calculated based on the target risk information and the predicted risk information, and the initial risk prediction model is adjusted according to the loss value until a risk prediction model that meets the training stop conditions is obtained.

[0137] The device for training a risk prediction model based on cross-domain text provided in this embodiment is to be able to train a model that can replace manual risk assessment for a part change task, so as to improve the real-time and accuracy of risk assessment. The global text data can be read from the global data source associated with the part change task first, and the local text data matched by the global text data can be read from the local data source associated with the part change task; the richness and universality of the samples can be improved by using the data from two different data sources and aligned as samples for training. Thereafter, the global text data and the local text data are input into the initial risk prediction model, and the global text data and the local text data are fused through the fusion module in the initial risk prediction model; and the fusion result is processed by the prediction module in the initial risk prediction model, and the predicted risk information output by the initial risk prediction model can be obtained according to the processing result; on this basis, the initial risk prediction model can be adjusted based on the target risk information and predicted risk information corresponding to the global text data, until a risk prediction model that meets the training stop condition is obtained. In the model training stage, the cross-domain text can be selected to train the model in the current scene, so that the model prediction accuracy can be effectively improved. This enables real-time and efficient risk prediction in part change scenarios, thereby effectively reducing dependence on experts.

[0138] The above is a schematic scheme of a device based on a cross-domain text training risk prediction model of this embodiment. It should be noted that the technical scheme of the device based on a cross-domain text training risk prediction model and the technical scheme of the method based on a cross-domain text training risk prediction model belong to the same concept. The details of the technical scheme of the device based on a cross-domain text training risk prediction model that are not described in detail can be found in the description of the technical scheme of the method based on a cross-domain text training risk prediction model.

[0139] Corresponding to the above method embodiment, this specification also provides a risk prediction device embodiment, Figure 5 FIG. 1 is a schematic diagram showing the structure of a risk prediction device provided by an embodiment of the present specification. Figure 5 As shown, the device comprises:

[0140] The data acquisition module 502 is configured to acquire part description data associated with the target part change task;

[0141] The model prediction module 504 is configured to input the part description data into the risk prediction model in the above method for processing to obtain part risk information;

[0142] The update information module 506 is configured to update the prompt information corresponding to the target part change task according to the part risk information, and display the updated prompt information.

[0143] In summary, by using risk prediction models instead of manually configuring rules to complete risk prediction of changed parts, the efficiency and accuracy of risk prediction can be effectively improved, and the consumption of human resources can be reduced to ensure that risk detection can be completed quickly in part change tasks and reduce the impact of risk events.

[0144] The above is a schematic scheme of a risk prediction device of this embodiment. It should be noted that the technical scheme of the risk prediction device and the technical scheme of the risk prediction method described above belong to the same concept, and the details of the technical scheme of the risk prediction device that are not described in detail can be found in the description of the technical scheme of the risk prediction method described above.

[0145] Figure 6 The block diagram of a computing device 600 according to an embodiment of the present specification is shown. The components of the computing device 600 include but are not limited to a memory 610 and a processor 620. The processor 620 is connected to the memory 610 via a bus 630, and the database 650 is used to store data.

[0146] The computing device 600 also includes an access device 640 that enables the computing device 600 to communicate via one or more networks 660. Examples of these networks include a public switched telephone network (PSTN), a local area network (LAN), a wide area network (WAN), a personal area network (PAN), or a combination of communication networks such as the Internet. The access device 640 may include one or more of any type of network interface (e.g., a network interface card (NIC)) of wired or wireless, such as an IEEE 802.11 wireless local area network (WLAN) wireless interface, a world-wide interoperability for microwave access (Wi-MAX) interface, an Ethernet interface, a universal serial bus (USB) interface, a cellular network interface, a Bluetooth interface, and a near field communication (NFC).

[0147] In one embodiment of the present specification, the above components of the computing device 600 and Figure 6Other components not shown in the figure may also be connected to each other, for example, via a bus. It should be understood that Figure 6 The computing device structure block diagram shown is only for the purpose of illustration, and is not intended to limit the scope of this specification. Those skilled in the art can add or replace other components as needed.

[0148] The computing device 600 may be any type of stationary or mobile computing device, including a mobile computer or mobile computing device (e.g., a tablet computer, a personal digital assistant, a laptop computer, a notebook computer, a netbook, etc.), a mobile phone (e.g., a smart phone), a wearable computing device (e.g., a smart watch, smart glasses, etc.), or other types of mobile devices, or a stationary computing device such as a desktop computer or a personal computer (PC). The computing device 600 may also be a mobile or stationary server.

[0149] Among them, the processor 620 is used to execute the following computer-executable instructions, which, when executed by the processor, implement the above-mentioned method for training a risk prediction model based on cross-domain text or the steps of the risk prediction method.

[0150] The above is a schematic scheme of a computing device of this embodiment. It should be noted that the technical scheme of the computing device and the technical scheme of the risk prediction model based on cross-domain text training or the risk prediction method are of the same concept. For details not described in detail in the technical scheme of the computing device, please refer to the description of the technical scheme of the risk prediction model based on cross-domain text training or the risk prediction method.

[0151] An embodiment of the present specification also provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the above-mentioned method for training a risk prediction model based on cross-domain text or the steps of the risk prediction method.

[0152] The above is a schematic scheme of a computer-readable storage medium of this embodiment. It should be noted that the technical scheme of the storage medium and the technical scheme of the risk prediction model based on cross-domain text training or the risk prediction method belong to the same concept, and the details not described in detail in the technical scheme of the storage medium can be found in the description of the technical scheme of the risk prediction model based on cross-domain text training or the risk prediction method.

[0153] An embodiment of the present specification also provides a computer program, wherein when the computer program is executed in a computer, the computer is caused to execute the above-mentioned method for training a risk prediction model based on cross-domain text or the steps of the risk prediction method.

[0154] The above is a schematic scheme of a computer program of this embodiment. It should be noted that the technical scheme of the computer program and the technical scheme of the risk prediction model based on cross-domain text training or the risk prediction method are of the same concept, and the details not described in detail in the technical scheme of the computer program can be found in the description of the technical scheme of the risk prediction model based on cross-domain text training or the risk prediction method.

[0155] The above is a description of a specific embodiment of the specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in an order different from that in the embodiments and still achieve the desired results. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0156] The computer instructions include computer program codes, which may be in source code form, object code form, executable files or some intermediate forms, etc. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal and software distribution medium, etc. It should be noted that the content contained in the computer-readable medium may be appropriately increased or decreased according to the requirements of patent practice. For example, in some regions, according to patent practice, computer-readable media do not include electric carrier signals and telecommunication signals.

[0157] It should be noted that, for the convenience of description, the aforementioned method embodiments are all described as a series of action combinations, but those skilled in the art should be aware that the embodiments of this specification are not limited by the order of the actions described, because according to the embodiments of this specification, some steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the embodiments of this specification.

[0158] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0159] The preferred embodiments of this specification disclosed above are only used to help explain this specification. The optional embodiments do not describe all the details in detail, nor do they limit the invention to the specific implementation methods described. Obviously, many modifications and changes can be made according to the content of the embodiments of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the embodiments of this specification, so that technicians in the relevant technical field can understand and use this specification well.

Claims

1. A method for training a risk prediction model based on cross-domain text, characterized in that: include: Reading global text data from a global data source associated with a part change task, and reading local text data matching the global text data from a local data source associated with the part change task; Inputting the global text data and the local text data into an initial risk prediction model, and fusing the global text data and the local text data through a fusion module in the initial risk prediction model; Processing the fusion result through the prediction module in the initial risk prediction model to obtain the predicted risk information output by the initial risk prediction model; Based on the target risk information and the predicted risk information corresponding to the global text data, the initial risk prediction model is adjusted until a risk prediction model that meets the training stop condition is obtained.

2. The method according to claim 1, characterized in that The reading of global text data from a global data source associated with the part change task, and the reading of local text data matched by the global text data from a local data source associated with the part change task, include: Determine a global data source and a local data source associated with the part change task, wherein the global data source is used to record text data associated with all part change processes in the part change task, and the local data source is used to record text data associated with part change anomalies in all part change processes; The global text data is read from the global data source, and the local data source is searched according to the global text data to obtain the local text data matching the global text data.

3. The method according to claim 2, characterized in that The step of searching the local data source according to the global text data to obtain the local text data matching the global text data includes: Searching the local data source according to the global text data, and detecting whether there is text data matching the global text data in the local data source according to the search result; If so, reading text data matching the global text data from the local data source as the local text data; If not, reference text data is read from the local data source as the local text data.

4. The method according to claim 1, characterized in that The fusing of the global text data and the local text data by a fusion module in the initial risk prediction model includes: The global text data and the local text data are respectively encoded by the encoding unit included in the fusion module in the initial risk prediction model to obtain a global text vector corresponding to the global text data and a local text vector corresponding to the local text data; The global text vector and the local text vector are fused by a fusion unit included in the fusion module to obtain a text fusion vector; The step of processing the fusion result by the prediction module in the initial risk prediction model to obtain the predicted risk information output by the initial risk prediction model includes: The text fusion vector is processed by the prediction module in the initial risk prediction model to obtain the predicted risk information output by the initial risk prediction model.

5. The method according to claim 4, characterized in that The determination of the global text vector includes: In the case where the global text data is a discrete data type, encoding the global text data by a first sub-encoding unit in the encoding unit to obtain the global text vector; In the case where the global text data is a continuous data type, the global text data is converted into target global text data of a corresponding discrete data type, and the target global text data is encoded by a second sub-encoding unit in the encoding unit to obtain the global text vector.

6. The method according to claim 4, characterized in that The determination of the local text vector includes: By using the word segmentation unit in the encoding unit, the local text data is segmented according to a pre-generated word list to obtain a plurality of word units; The plurality of word units are encoded by a third sub-encoding unit in the encoding unit to obtain the local text vector.

7. The method according to any one of claims 1 to 6, characterized in that: The step of adjusting the parameters of the initial risk prediction model based on the target risk information corresponding to the global text data and the predicted risk information until a risk prediction model that meets the training stop condition is obtained includes: Detecting the global text data according to the business logic rules corresponding to the part change task, and determining the target risk information corresponding to the global text data according to the detection result; A loss value is calculated based on the target risk information and the predicted risk information, and the initial risk prediction model is adjusted according to the loss value until a risk prediction model that meets the training stop condition is obtained.

8. A risk prediction method, characterized in that: include: Obtain part description data associated with the target part change task; Inputting the part description data into the risk prediction model in the method according to any one of claims 1 to 7 for processing to obtain part risk information; The prompt information corresponding to the target part change task is updated according to the part risk information, and the updated prompt information is displayed.

9. A device for training a risk prediction model based on cross-domain text, characterized in that: include: A reading module, configured to read global text data from a global data source associated with a part change task, and to read local text data matching the global text data from a local data source associated with the part change task; An input module, configured to input the global text data and the local text data into an initial risk prediction model, and fuse the global text data and the local text data through a fusion module in the initial risk prediction model; An output module is configured to process the fusion result through the prediction module in the initial risk prediction model to obtain the predicted risk information output by the initial risk prediction model; The parameter adjustment module is configured to adjust the parameters of the initial risk prediction model based on the target risk information and the predicted risk information corresponding to the global text data until a risk prediction model that meets the training stop condition is obtained.

10. A risk prediction device, characterized in that: include: A data acquisition module is configured to acquire part description data associated with a target part change task; A model prediction module, configured to input the part description data into the risk prediction model in the method according to any one of claims 1 to 7 for processing to obtain part risk information; The update information module is configured to update the prompt information corresponding to the target part change task according to the part risk information, and display the updated prompt information.

11. A computing device, characterized in that: include: Memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, the steps of the method described in any one of claims 1 to 8 are implemented.

12. A computer-readable storage medium, characterized in that: It stores computer executable instructions, which, when executed by a processor, implement the steps of the method described in any one of claims 1 to 8.

Images