Work order early warning method, device and electronic equipment

Abstract

An embodiment of the present invention provides a work order early warning method, device and electronic device, the method comprising: determining the work order fault content, processing notes, work order processing time limit and processing link type of a target work order, wherein the work order fault content and the processing notes are text information, and the work order processing time limit and processing link type are digital information; inputting the work order fault content, processing notes, work order processing time limit and processing link type into a first neural network model to obtain a first work order early warning degree output by the first neural network model; if the first work order early warning degree is greater than a preset threshold, issuing an early warning, thereby reducing the risk of work order timeout.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a work order early warning method, device and electronic equipment. Background Art

[0002] Currently, when processing customer fault work orders, each work order type is different and the links are different. If a relative timeout occurs in one link, it may cause the entire work order processing to time out. Each work order link requires manual control. Due to the large number of work orders, a large number of people are often required, and there is often a shortage of staff. In addition, manual control may also result in errors, which in turn leads to further timeouts in work order processing. Summary of the invention

[0003] The purpose of the embodiments of the present invention is to provide a work order early warning method, device and electronic device to reduce the risk of work order timeout. The specific technical solution is as follows:

[0004] In a first aspect of the present invention, a work order early warning method is provided, the method comprising:

[0005] Determine the work order fault content, processing notes, work order processing time limit and processing link type of the target work order, wherein the work order fault content and the processing notes are text information, and the work order processing time limit and the processing link type are digital information;

[0006] Inputting the work order fault content, the processing remarks, the work order processing time limit and the processing link type into a first neural network model to obtain a first work order warning level output by the first neural network model;

[0007] If the warning level of the first work order is greater than a preset threshold, a warning is issued;

[0008] Wherein, the first neural network model is pre-trained with sample data, and the sample data includes:

[0009] The sample work order includes the work order failure content, processing notes, work order processing time limit, processing link type and work order warning information, wherein the work order warning information is used to indicate the timeout status of the sample work order.

[0010] In a possible embodiment, the first neural network model includes: a bidirectional recurrent neural network, a unidirectional recurrent neural network, and a deep neural network, wherein the bidirectional recurrent neural network and the unidirectional recurrent neural network are used to process textual information in the input of the first neural network model, and the deep neural network is used to process numerical information in the input of the first neural network model.

[0011] In a possible embodiment, the inputting the work order fault content, the processing remarks, the work order processing time limit and the processing link type into a first neural network model to obtain a first work order warning level output by the first neural network model includes:

[0012] Segmenting the work order fault content and the processing remarks to obtain work order fault content segmentation and processing remarks segmentation;

[0013] Input the work order fault content segmentation and the processing note description segmentation into a pre-trained word library to obtain a work order fault content index and a processing note description index;

[0014] Convert the work order fault content index and the processing note description index into a work order fault content word vector and a processing note description word vector;

[0015] The work order fault content word vector, processing note description word vector, work order processing time limit and link status type are input into the first neural network model to obtain the first work order warning level output by the first neural network model.

[0016] In a possible embodiment, the work order fault content, the processing remarks, the work order processing time limit and the processing link type are input into a first neural network model to obtain a first work order warning level output by the first neural network model, including:

[0017] The work order fault content, the processing notes, the work order processing time limit and the processing link type are input into the first neural network model, so that the first neural network model weightedly stacks the work order processing time limit with the link status type and the first learnable weight and the second learnable weight to obtain a third learnable weight, and obtain the first work order warning level output by the first neural network model, wherein the first learnable weight is the weight corresponding to the work order processing time limit, the second learnable weight is the weight corresponding to the link status type, the fourth learnable weight is the weight corresponding to the work order fault content, the fifth learnable weight is the weight corresponding to the processing notes, the first learnable weight is negatively correlated with the second learnable weight, and the third learnable weight, the fourth learnable weight and the fifth learnable weight are negatively correlated.

[0018] In a possible embodiment, the first neural network model is pre-trained in the following manner:

[0019] Setting a first learnable weight for the work order processing time limit of the sample work order, and obtaining a second learnable weight according to the first learnable weight;

[0020] Inputting the work order fault content, processing remarks, work order processing time limit, processing link type and work order warning information of the sample work order into the initial neural network model, so that the initial neural network model outputs the second work order warning level;

[0021] The first learnable weight, the second learnable weight, the third learnable weight, the fourth learnable weight and the fifth learnable weight are adjusted according to the second work order warning level and the work order warning information to obtain the first neural network model.

[0022] In a possible embodiment, obtaining the first neural network model includes:

[0023] Inputting the work order fault content, processing remarks, work order processing time limit, processing link type and work order warning information of the verification work order into the initial neural network model with adjusted weights, so that the initial neural network model with adjusted weights outputs the third work order warning degree;

[0024] Verifying the warning level of the third work order according to the work order warning information;

[0025] If the third work order warning level meets the work order warning information, using the initial neural network model with adjusted weights as the first neural network model;

[0026] If the third work order warning level does not conform to the work order warning information, the initial neural network with adjusted weights is retrained.

[0027] In a second aspect of the present invention, a work order early warning device is provided, the device comprising:

[0028] A data acquisition module, used to determine the work order fault content, processing notes, work order processing time limit and processing link type of the target work order, wherein the work order fault content and the processing notes are text information, and the work order processing time limit and the processing link type are digital information;

[0029] A warning degree acquisition module, used for inputting the work order fault content, the processing remarks, the work order processing time limit and the processing link type into the first neural network model to obtain the first work order warning degree output by the first neural network model;

[0030] An early warning module, configured to issue an early warning if the early warning level of the first work order is greater than a preset threshold;

[0031] Wherein, the first neural network model is pre-trained with sample data, and the sample data includes:

[0032] The sample work order includes the work order failure content, processing notes, work order processing time limit, processing link type and work order warning information, wherein the work order warning information is used to indicate the timeout status of the sample work order.

[0033] In a possible embodiment, the first neural network model includes: a bidirectional recurrent neural network, a unidirectional recurrent neural network, and a deep neural network, wherein the bidirectional recurrent neural network and the unidirectional recurrent neural network are used to process textual information in the input of the first neural network model, and the deep neural network is used to process numerical information in the input of the first neural network model.

[0034] In a possible embodiment, the warning level acquisition module includes:

[0035] A word segmentation unit is used to segment the work order fault content and the processing remarks to obtain work order fault content segmentation and processing remarks segmentation;

[0036] An indexing unit, used to input the work order fault content segmentation and the processing note description segmentation into a pre-trained word library to obtain a work order fault content index and a processing note description index;

[0037] A word vector acquisition unit, used to convert the work order fault content index and the processing note description index into a work order fault content word vector and a processing note description word vector;

[0038] The warning degree acquisition unit is used to input the work order fault content word vector, processing note description word vector, work order processing time limit and link status type into the first neural network model to obtain the first work order warning degree output by the first neural network model.

[0039] In a possible embodiment, the warning level acquisition module is specifically used to input the work order fault content, the processing notes, the work order processing time limit and the processing link type into a first neural network model, so that the first neural network model weightedly stacks the work order processing time limit with the link status type and the first learnable weight and the second learnable weight to obtain a third learnable weight, and obtain the first work order warning level output by the first neural network model, wherein the first learnable weight is the weight corresponding to the work order processing time limit, the second learnable weight is the weight corresponding to the link status type, the fourth learnable weight is the weight corresponding to the work order fault content, the fifth learnable weight is the weight corresponding to the processing notes, the first learnable weight is negatively correlated with the second learnable weight, and the third learnable weight, the fourth learnable weight and the fifth learnable weight are negatively correlated.

[0040] In a possible embodiment, the device further includes:

[0041] A learning weight setting module, used to set a first learnable weight for the work order processing time limit of the sample work order, and obtain a second learnable weight according to the first learnable weight;

[0042] A second work order warning acquisition module is used to input the work order fault content, processing remarks, work order processing time limit, processing link type and work order warning information of the sample work order into the initial neural network model, so that the initial neural network model outputs the second work order warning level;

[0043] A neural network model acquisition module is used to adjust the first learnable weight, the second learnable weight, the third learnable weight, the fourth learnable weight and the fifth learnable weight according to the second work order warning level and the work order warning information to obtain the first neural network model.

[0044] In a possible embodiment, the neural network model acquisition module includes:

[0045] A third work order warning acquisition unit is used to input the work order fault content, processing remarks, work order processing time limit, processing link type and work order warning information of the verification work order into the initial neural network model with adjusted weights, so that the initial neural network model with adjusted weights outputs a third work order warning degree;

[0046] A verification unit, configured to verify the warning level of the third work order according to the work order warning information;

[0047] A neural network model acquisition unit, configured to use the initial neural network model with adjusted weights as the first neural network model if the third work order warning level meets the work order warning information;

[0048] A training unit is used to retrain the initial neural network with adjusted weights if the third work order warning level does not conform to the work order warning information.

[0049] In a third aspect of the present invention, the present invention further provides an electronic device, comprising a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other via the communication bus;

[0050] Memory, used to store computer programs;

[0051] The processor is used to implement any method step described in the first aspect when executing the program stored in the memory.

[0052] In a fourth aspect of the present invention, the present invention further provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method steps described in any one of the first aspects are implemented.

[0053] Beneficial effects of the embodiments of the present invention:

[0054] A work order warning method provided by an embodiment of the present invention determines the work order fault content, processing remarks, work order processing time limit and processing link type of a target work order, and inputs each key information about the target work order into a trained first neural network model, thereby extracting key features that affect the processing time limit of the target work order, so that the first neural network model outputs a first work order warning degree. When the first work order warning degree exceeds a preset threshold, a timeout warning is issued for the target work order, thereby realizing automatic warning of timeout of the target work order, so that corresponding measures can be taken for the timed-out target work order later, thereby reducing the timeout risk of the work order.

[0055] Of course, it is not necessary to achieve all of the advantages described above at the same time to implement any product or method of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention. For ordinary technicians in this field, other embodiments can also be obtained based on these drawings.

[0057] Figure 1 A schematic diagram of a process flow of a work order early warning method provided by an embodiment of the present invention;

[0058] Figure 2 A schematic diagram of the structure of a first neural network model provided by an embodiment of the present invention;

[0059] Figure 3 A schematic diagram of a process flow of another work order early warning method provided by an embodiment of the present invention;

[0060] Figure 4 A schematic diagram of a process flow of another work order early warning method provided by an embodiment of the present invention;

[0061] Figure 5 A partial structural diagram of a first neural network model provided by an embodiment of the present invention;

[0062] Figure 6 Another partial structural diagram of the first neural network model provided by an embodiment of the present invention;

[0063] Figure 7 A schematic diagram of a flow chart of a training method for a first neural network model provided in an embodiment of the present invention;

[0064] Figure 8 A schematic flow chart of another first neural network model training method provided by an embodiment of the present invention;

[0065] Fig. 9 A schematic diagram of the structure of a work order early warning device provided by an embodiment of the present invention;

[0066] Fig.10 A schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0067] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field based on the present invention belong to the scope of protection of the present invention.

[0068] See also Figure 1 , Figure 1 A flowchart of a work order early warning method provided by an embodiment of the present invention may include:

[0069] S101, determine the work order fault content, processing remarks, work order processing time limit and processing link type of the target work order.

[0070] S102, inputting the work order fault content, processing remarks, work order processing time limit and processing link type into the first neural network model to obtain the first work order warning level output by the first neural network model.

[0071] S103: If the warning level of the first work order is greater than a preset threshold, issue a warning.

[0072] This embodiment is selected to determine the work order fault content, processing notes, work order processing time limit and processing link type of the target work order, and input the key information about the target work order into the trained first neural network model, so as to extract the key features that affect the processing time limit of the target work order, so that the first neural network model outputs the first work order warning level. When the first work order warning level exceeds the preset threshold, the target work order is given a timeout warning, thereby realizing automatic warning of target work order timeout, so that corresponding measures can be taken for the timed target work order later, thereby reducing the timeout risk of the work order.

[0073] Among them, in S101, the work order fault content and processing notes are text information represented by text, and the work order processing time limit and processing link type are digital information represented by numerical values. The work order fault content can be the fault reported by the customer contained in the target work order, and the processing notes can be the special processing requirements of the target work order. It can be understood that the work order fault content and processing notes are usually expressed through text information. The work order processing time limit can include the processing time limit of the entire target work order and the processing time limit of each processing link in the target work order. The processing link type can be the processing status of each link in the target work order. For example, the processed link can be identified as 1, the unprocessed link can be identified as 0, and the link being processed can be identified as 2. It can be understood that the above-mentioned work order processing time limit and processing link type can be represented by digital information. When receiving a fault work order from a customer, the work order usually includes the work order title, the fault content of the work order, the processing time limit of the work order, the processing notes of the work order, the processing link type of the work order and the link time of each link in the work order, etc. From the various information of the above work orders, the work order fault content, processing notes, work order processing time limit and processing link type of the target work order can be obtained. The content used to represent the work order fault content, processing notes, work order processing time limit and processing link type in the target work order has certain traceable time sequence characteristics and contains high-level semantic features. Therefore, the work order fault content, processing notes, work order processing time limit and processing link type of the target work order can be obtained to facilitate the subsequent extraction of various features that can reflect whether the target work order has timed out.

[0074] Among them, in S102, the first neural network model is pre-trained with sample data, and the sample data includes: the work order fault content of the sample work order, processing notes, work order processing time limit, processing link type and work order warning information, among which the work order warning information is used to indicate the timeout status of the sample work order.

[0075] The first neural network model is pre-trained with sample data. In addition to the work order fault content, processing notes, work order processing time limit and processing link type of the sample work order, the sample data also includes work order warning information used to indicate the timeout situation of the sample work order. It can be understood that the work order warning information can be used to indicate whether the processing of the sample work order has timed out. Since the sample data contains the work order warning information of the sample work order, in the subsequent process of using the sample data to train the initial neural network model to obtain the first neural network model, the initial neural network model can be adjusted according to the output work order warning level of the sample work order and the actual work order warning information of the sample work order to obtain the first neural network model, so that the first neural network model can accurately predict the first work order warning level of the target work order.

[0076] The work order fault content, processing notes, work order processing time limit and processing link type in the target work order are input into the trained first neural network model. The first neural network model can be used to extract the various features of the target work order, thereby obtaining the first work order warning level that can accurately reflect the timeout situation of the target work order. It can be understood that the first work order warning level can be the probability that the total processing time limit of the target work order will exceed the specified time limit.

[0077] Among them, in S103, the preset threshold can be pre-set by technical personnel in this field based on experience and / or the training situation of the first neural network model. The preset threshold can be a threshold expressed in the form of probability, such as 50%, 70%, etc. In other possible embodiments, the preset threshold can also be expressed in a form other than probability. If the probability represented by the warning level of the first work order is greater than the preset threshold, it is judged that the processing of the target work order will time out, and it is necessary to warn the target work order so that the target work order can be further accelerated.

[0078] In actual applications, although the data dimensions of the target work order, such as the work order fault content, processing notes, work order processing time limit and processing link type, contain high-level semantic features, these features cannot be extracted based on traditional machine learning. Therefore, the present invention also provides a method for constructing a first neural network model. In a possible embodiment, the first neural network model includes: a bidirectional recurrent neural network, a unidirectional recurrent neural network, and a deep neural network.

[0079] Among them, the bidirectional recurrent neural network and the unidirectional recurrent neural network are used to process text information in the input of the first neural network model, and the deep neural network is used to process digital information in the input of the first neural network model.

[0080] As mentioned above, among the data of the work order fault content, processing notes, work order processing time limit and processing link type obtained for the target work order, the work order fault content and processing notes are text information, and the work order processing time limit and processing link type are digital information. Therefore, it is impossible to use a simple single-process neural network to process the above data like the traditional neural network model, but it is necessary to use a multi-input hybrid neural network structure of a recurrent neural network (RNN) and a deep neural network (DNN) to build a neural network model. Among them, the recurrent neural network can include a bidirectional recurrent neural network and a unidirectional recurrent neural network, both of which are used to process text information such as work order fault content and processing notes. Specifically, the bidirectional recurrent neural network can not only obtain the information before a certain point in the text information sequence, but also predict the information after the point, while the unidirectional recurrent neural network can process sequences of any length, and is widely used in natural language processing, such as machine translation, text generation, question-answering system, text classification, etc. Combining the bidirectional recurrent neural network and the unidirectional recurrent neural network to process the text information in the first neural network model can well extract the characteristics of the text information expression. Deep neural networks can represent complex functions with fewer parameters and can process simple digital information very well.

[0081] In a possible embodiment, Figure 2 As shown in the structural diagram of the first neural network model, the bidirectional long short-term memory network (Bi-LSTM) can be used as the network of the bidirectional recurrent neural network layer, and the unidirectional long short-term memory network (LSTM) can be used as the network of the unidirectional recurrent neural network layer. The bidirectional long short-term memory network is composed of two unidirectional long short-term memory network recurrent layers with opposite information transmission, in which the first layer transmits information in chronological order and the second layer transmits information in reverse chronological order. In this way, not only past information can be utilized, but also subsequent information can be captured. For example, in the feature acquisition of text information, the bidirectional long short-term memory network can make good use of context information. The unidirectional long short-term memory network can retain past information. Using the unidirectional long short-term memory network after the bidirectional long short-term memory network can retain the features of the text information that has been acquired, further improving the efficiency of text information feature acquisition.

[0082] By selecting this embodiment, a targeted first neural network model can be constructed for the data information such as work order fault content, processing notes, work order processing time limit and processing link type, so as to accurately and effectively extract the semantic features contained therein, thereby enabling the first neural network model to accurately predict the first work order warning level of the target work order.

[0083] In a possible embodiment, after the work order fault content, processing notes, work order processing time limit, and processing link type are processed by a bidirectional recurrent neural network, a unidirectional recurrent neural network, and a deep neural network, they can also be further processed by a fully connected network. Specifically, the fully connected network can use a Sigmoid function as an activation function.

[0084] It is understandable that text information is different from digital information. If the text information is directly input into the first neural network model, the model cannot directly recognize and calculate it. Therefore, the present invention also provides a work order warning method, such as Figure 3 As shown, including:

[0085] S301, determine the work order fault content, processing remarks, work order processing time limit and processing link type of the target work order.

[0086] This step is the same as the aforementioned S101. Please refer to the relevant description of the aforementioned S101 and will not be repeated here.

[0087] S302, segmenting the work order fault content and the processing remarks to obtain the work order fault content segmentation and the processing remarks segmentation.

[0088] S303, input the work order fault content segmentation and the processing note description segmentation into a pre-trained word library to obtain the work order fault content index and the processing note description index.

[0089] S304, converting the work order fault content index and the processing note description index into the work order fault content word vector and the processing note description word vector.

[0090] S305, input the word vector of the work order fault content, the word vector of the processing remarks, the work order processing time limit and the link status type into the first neural network model to obtain the first work order warning level output by the first neural network model.

[0091] S306: If the warning level of the first work order is greater than a preset threshold, issue a warning.

[0092] This step is the same as the aforementioned S103. Please refer to the relevant description of the aforementioned S103 and will not be repeated here.

[0093] Among them, in S302, the work order fault content and processing notes are segmented, and the work order fault content and processing notes are segmented according to the subject-predicate-object grammar. For example, "I like literature" is segmented to obtain "I, like, literature". Specifically, a segmentation tool such as Jieba Segmentation can be used for word segmentation.

[0094] In a possible embodiment, a word segmentation library containing 260,000 word vectors open sourced by researchers from the Institute of Chinese Information Processing of Beijing Normal University and the DBIIR Laboratory of Renmin University of China can be used to segment the work order fault content and processing notes, index them and obtain word vectors, thereby improving the accuracy of the obtained word vectors.

[0095] Among them, in S303, after obtaining the work order fault content segmentation and the processing note segmentation, the above segmentation can be input into the pre-trained word library, and the index identifier for each segmentation is determined by the pre-trained word library, and the corresponding index identifiers are arranged in the text order of the work order fault content and the processing note description, so as to obtain the work order fault content index and the processing note description index. Exemplarily, if the work order fault content segmentation obtained is "I, like, literature", and the index identifier of "I" is determined to be 2, the index identifier of "like" is 345, and the index identifier of "literature" is 4564 in the pre-trained word library, then it can be determined that the index corresponding to "I like literature" is [2,345,4564].

[0096] Among them, in S304, the obtained work order fault content index and processing notes index are vectorized, that is, the above index is converted into digital information in vector form. Exemplarily, the process of converting the above index into a word vector can be to convert the above index [2,345,4564] with a vector length of 3 into a matrix with a dimension of 300 and a length of 3, so that a word vector represented in matrix form is obtained. It can be understood that the dimension of the word vector can be set according to needs, and the present invention is not limited to this.

[0097] Among them, in S305, text information has been converted into word vectors that can be recognized by the first neural network model, and digital information can be directly input into the first neural network model. Therefore, in this step, the work order fault content word vector, processing notes description word vector, work order processing time limit and link status type are input into the first neural network model.

[0098] By selecting this embodiment, text information can be accurately converted into word vectors that can be recognized by the first neural network model, ensuring the normal reception of input by the first neural network model, and further improving the accuracy of feature extraction of input text information by the first neural network model.

[0099] In a possible embodiment, the present invention further provides a work order early warning method, such as Figure 4 As shown, the method includes:

[0100] S401, determine the work order fault content, processing remarks, work order processing time limit and processing link type of the target work order.

[0101] This step is the same as the aforementioned S101. Please refer to the relevant description of the aforementioned S101 and will not be repeated here.

[0102] S402, input the work order fault content, processing notes, work order processing time limit and processing link type into the first neural network model, so that the first neural network model weightedly stacks the work order processing time limit and link status type and the first learnable weight and the second learnable weight to obtain the third learnable weight, and obtain the first work order warning level output by the first neural network model.

[0103] S403: If the warning level of the first work order is greater than a preset threshold, issue a warning.

[0104] This step is the same as the aforementioned S103. Please refer to the relevant description of the aforementioned S103 and will not be repeated here.

[0105] Among them, in S402, the first learnable weight is the weight corresponding to the work order processing time limit, the second learnable weight is the weight corresponding to the link status type, the fourth learnable weight is the weight corresponding to the work order fault content, and the fifth learnable weight is the weight corresponding to the processing notes. The first learnable weight is negatively correlated with the second learnable weight, and the third learnable weight and the fourth learnable weight are negatively correlated with the fifth learnable weight.

[0106] After the first neural network model obtains the features corresponding to the work order fault content, the processing notes, the work order processing time limit and the processing link type, it is also necessary to determine the weights for the above features so that the above features can be weighted later, so that the first neural network model can output the first work order warning level of the target work order. Specifically, the work order fault content corresponds to the fourth learnable weight, the processing notes correspond to the fifth learnable weight, the work order processing time limit corresponds to the first learnable weight, and the processing link type corresponds to the second learnable weight. Since the work order processing time limit and the processing link type are both digital information, the work order processing time limit and the processing link type input to the first neural network model are processed together, and the first learnable weight and the second learnable weight are used as a group of weights, so that the first learnable weight is negatively correlated with the second learnable weight. Exemplarily, the sum of the first learnable weight and the second learnable weight can be set to 1. The weights corresponding to the work order processing time limit and the processing link type are weighted and superimposed, so as to obtain the third learnable weight corresponding to all digital information. Then the digital information is associated with the textual information, and the third learnable weight, the fourth learnable weight and the fifth learnable weight are negatively correlated with each other. Exemplarily, the sum of the third learnable weight, the fourth learnable weight and the fifth learnable weight can be set to 1.

[0107] By selecting this embodiment, a reasonably allocated weight can be obtained for each feature, thereby improving the accuracy of the obtained first work order warning level for predicting the target work order.

[0108] In a possible embodiment, the Sigmoid function can be used as the weight control function of the first learnable function and the second learnable weight, see Figure 5 The partial structural diagram of the first neural network model shown in FIG. 1 , specifically, the first learnable weight can be set to θ1, the second learnable weight can be set to θ2, the work order processing time limit can be set to x1, the processing link type can be set to x2, and the result of the weighted superposition of the work order processing time limit and the processing link type can be set to θ3, then:

[0109]

[0110] θ2=1-θ1

[0111] θ3=Sigmoid(θ1)×x1+x2×(1-Sigmoid(θ1))×Sigmoid(θ2)

[0112] In a possible embodiment, a Softmax function may be used as a weight control function of the fourth learnable function and the fifth learnable weight, see Figure 6 Partial structure diagram of the first neural network model shown in FIG. 1 , specifically, the fourth learnable weight can be set to θ4, the fifth learnable weight can be set to θ5, and the output of the previous node can be set to z1, z2, and z3 respectively, then

[0113]

[0114]

[0115]

[0116] Thus, we obtain θ3, θ4, and θ5, which need to be superimposed.

[0117] It is understandable that the present invention sets weights for each feature in the first neural network model, but the optimal weight setting cannot be obtained by setting it only based on manual experience, which will affect the accuracy of the first work order warning level output by the first neural network. Based on this, the present invention also provides a training method for the first neural network model, such as Figure 7 As shown, including:

[0118] S701, setting a first learnable weight for a work order processing time limit of a sample work order, and obtaining a second learnable weight according to the first learnable weight.

[0119] S702, input the work order fault content, processing remarks, work order processing time limit, processing link type and work order warning information of the sample work order into the initial neural network model, so that the initial neural network model outputs the second work order warning level.

[0120] S703, adjusting the first learnable weight, the second learnable weight, the third learnable weight, the fourth learnable weight and the fifth learnable weight according to the second work order warning level and the work order warning information to obtain a first neural network model.

[0121] Among them, in S701, the weight corresponding to the work order processing time limit of the sample work order is set as the first learnable weight. Specifically, the value of the first learnable weight can be set according to the experience of technical personnel in this field, and the first learnable weight is negatively correlated with the second learnable weight. Therefore, the second learnable weight can be obtained based on the set first learnable weight.

[0122] Among them, in S702, it can be understood that the structures of the first neural network model and the initial neural network model are the same, except that the initial neural network model is a model that has not been trained. The training data in the sample work order is input into the untrained initial neural network model, so that the initial neural network model processes each feature according to the set weights to obtain the second work order warning level predicted for the sample work order.

[0123] Among them, in S703, the second work order warning level is the timeout situation predicted by the initial neural network model for the sample work order, and the work order warning information of the sample work order includes the actual timeout situation of the sample work order. The learnable weights are adjusted according to the work order warning information and the second work order warning level. It can be to construct a target loss function about the difference between the second work order warning level and the work order warning information, and calculate the weight value that minimizes the function value of the target loss function through the back propagation algorithm, so as to optimize the first learnable weight set in advance according to the calculated weight value, and then obtain the adjusted second learnable weight, third learnable weight, fourth learnable weight and fifth learnable weight.

[0124] By selecting this embodiment, the pre-manually set weights can be adjusted to optimal weight values ​​through training, thereby assigning reasonable weight values ​​to the features obtained from the work order fault content, processing notes, work order processing time limit, and processing link type, so that the trained first neural network model can output an accurate first work order warning level.

[0125] In a possible embodiment, after the initial neural network model is trained and adjusted, if the adjusted initial neural network model is directly used as the first neural network model without verification, it may be impossible to know whether the first neural network model can accurately output the first work order warning level. Based on this, the present invention also provides a first neural network model training method, such as Figure 8 As shown, including:

[0126] S801, setting a first learnable weight for a work order processing time limit of a sample work order, and obtaining a second learnable weight according to the first learnable weight.

[0127] This step is the same as the aforementioned S701. Please refer to the relevant instructions of the aforementioned S701 and will not be repeated here.

[0128] S802, input the work order fault content, processing remarks, work order processing time limit, processing link type and work order warning information of the sample work order into the initial neural network model, so that the initial neural network model outputs the second work order warning level.

[0129] This step is the same as the aforementioned S702. Please refer to the relevant instructions of the aforementioned S702 and will not be repeated here.

[0130] S803 , adjusting the first learnable weight, the second learnable weight, the third learnable weight, the fourth learnable weight, and the fifth learnable weight according to the second work order warning level and the work order warning information.

[0131] This step is similar to the aforementioned S703, except that the initial neural network model after adjusting the weights is not directly used as the first neural network model. Please refer to the relevant instructions of the aforementioned S703, which will not be repeated here.

[0132] S804, input the work order fault content, processing notes, work order processing time limit, processing link type and work order warning information of the verification work order into the initial neural network model with adjusted weights, so that the initial neural network model with adjusted weights outputs the third work order warning level.

[0133] This step is similar to the aforementioned S702, except that the data of the verification work order is input, and the initial neural network model with adjusted weights is used. The warning level of the third work order corresponding to the verification work order is obtained. Please refer to the relevant instructions of the aforementioned S703, which will not be repeated here.

[0134] S805: Verify the warning level of the third work order according to the work order warning information.

[0135] S806: If the warning level of the third work order meets the warning information of the work order, the initial neural network model with adjusted weights is used as the first neural network model.

[0136] S807: If the third work order warning level does not meet the work order warning information, the initial neural network model with adjusted weights is retrained.

[0137] Among them, in S805, the work order warning information of the verification work order can be used to indicate whether the processing of the verification work order has timed out, and the third work order warning level can be the probability value of the verification work order timeout. Verifying the third work order warning level according to the work order warning information can be to determine whether the probability value in the third work order warning level is accurate. Exemplarily, if the work order warning information indicates that the processing of the verification work order is timed out, the probability value in the third work order warning level is greater than the preset threshold, that is, the third work order warning level predicts that the verification work order will time out. It can be understood that if the prediction of the third work order warning level is consistent with the actual situation, then the third work order warning level is consistent with the work order warning information. Conversely, if the probability value in the third work order warning level is less than the preset threshold, then the third work order warning level does not meet the work order warning information.

[0138] Among them, in S806, if the warning degree of the third work order is consistent with the warning information of the work order, it means that the initial neural network model with adjusted weights can accurately predict the timeout of the work order. Therefore, the initial neural network model with adjusted weights can be used as the first neural network model. In a possible embodiment, the initial neural network model with adjusted weights can be verified a preset number of times. If the results of the preset number of verifications all indicate that the warning degree of the third work order is consistent with the warning information of the work order, the initial neural network model with adjusted weights can be used as the first neural network model.

[0139] Among them, in S807, if the warning level of the third work order does not meet the work order warning information, it means that the initial neural network model with adjusted weights cannot accurately predict the timeout of the work order. Therefore, the initial neural network model needs to be further trained and the weights in the initial neural network model need to be adjusted again.

[0140] By selecting this embodiment, it can be ensured that the trained first neural network model accurately predicts the first work order warning level of the target work order, thereby improving the effectiveness of the first work order warning level.

[0141] In a possible embodiment, when training the first neural network model, the first neural network model can be constructed using the tensorflow 2.6 deep learning framework, and the DataLoader (data iterator) method is used to perform 50 epochs (1 epoch means that all data are sent to the network and a forward calculation and back propagation process is completed) in a small batch (mini-batch (small batch gradient descent), batch-size (batch size) = 64) mode.

[0142] In a possible embodiment, when training the first neural network model, the CUDA_VISIBLE_DEVICES (CUDA environment variable) environment parameter can be specified, wherein the CUDA (Unified Compute Device Architecture) environment variable CUDA_VISIBLE_DEVICES can be used to limit the GPUs that can be used by the CUDA program, thereby placing data separately on multiple GPU graphics cards for concurrent training, thereby improving training efficiency.

[0143] In a possible embodiment, when training the first neural network model, the input data dimension may also be adjusted so that the warning level of the first work order output by the first neural network model is more accurate.

[0144] See also Fig. 9 , Fig. 9 The figure is a schematic diagram of the structure of a work order early warning device provided by an embodiment of the present invention, which may include:

[0145] The data acquisition module 901 is used to determine the work order fault content, processing notes, work order processing time limit and processing link type of the target work order, wherein the work order fault content and the processing notes are text information, and the work order processing time limit and the processing link type are digital information;

[0146] The warning level acquisition module 902 is used to input the work order fault content, the processing remarks, the work order processing time limit and the processing link type into the first neural network model to obtain the first work order warning level output by the first neural network model;

[0147] An early warning module 903 is used to issue an early warning if the early warning level of the first work order is greater than a preset threshold;

[0148] Wherein, the first neural network model is pre-trained with sample data, and the sample data includes:

[0149] The sample work order includes the work order failure content, processing notes, work order processing time limit, processing link type and work order warning information, wherein the work order warning information is used to indicate the timeout status of the sample work order.

[0150] In a possible embodiment, the first neural network model includes: a bidirectional recurrent neural network, a unidirectional recurrent neural network, and a deep neural network, wherein the bidirectional recurrent neural network and the unidirectional recurrent neural network are used to process textual information in the input of the first neural network model, and the deep neural network is used to process numerical information in the input of the first neural network model.

[0151] In a possible embodiment, the warning level acquisition module 902 includes:

[0152] A word segmentation unit is used to segment the work order fault content and the processing remarks to obtain work order fault content segmentation and processing remarks segmentation;

[0153] An indexing unit, used to input the work order fault content segmentation and the processing note description segmentation into a pre-trained word library to obtain a work order fault content index and a processing note description index;

[0154] A word vector acquisition unit, used to convert the work order fault content index and the processing note description index into a work order fault content word vector and a processing note description word vector;

[0155] The warning degree acquisition unit is used to input the work order fault content word vector, processing note description word vector, work order processing time limit and link status type into the first neural network model to obtain the first work order warning degree output by the first neural network model.

[0156] In a possible embodiment, the warning level acquisition module 902 is specifically used to input the work order fault content, the processing notes, the work order processing time limit and the processing link type into the first neural network model, so that the first neural network model weightedly stacks the work order processing time limit with the link status type and the first learnable weight and the second learnable weight to obtain a third learnable weight, and obtain the first work order warning level output by the first neural network model, wherein the first learnable weight is the weight corresponding to the work order processing time limit, the second learnable weight is the weight corresponding to the link status type, the fourth learnable weight is the weight corresponding to the work order fault content, the fifth learnable weight is the weight corresponding to the processing notes, the first learnable weight is negatively correlated with the second learnable weight, and the third learnable weight, the fourth learnable weight and the fifth learnable weight are negatively correlated.

[0157] In a possible embodiment, the device further includes:

[0158] A learning weight setting module, used to set a first learnable weight for the work order processing time limit of the sample work order, and obtain a second learnable weight according to the first learnable weight;

[0159] A second work order warning acquisition module is used to input the work order fault content, processing remarks, work order processing time limit, processing link type and work order warning information of the sample work order into the initial neural network model, so that the initial neural network model outputs the second work order warning level;

[0160] A neural network model acquisition module is used to adjust the first learnable weight, the second learnable weight, the third learnable weight, the fourth learnable weight and the fifth learnable weight according to the second work order warning level and the work order warning information to obtain the first neural network model.

[0161] In a possible embodiment, the neural network model acquisition module includes:

[0162] A third work order warning acquisition unit is used to input the work order fault content, processing remarks, work order processing time limit, processing link type and work order warning information of the verification work order into the initial neural network model with adjusted weights, so that the initial neural network model with adjusted weights outputs a third work order warning degree;

[0163] A verification unit, configured to verify the warning level of the third work order according to the work order warning information;

[0164] A neural network model acquisition unit, configured to use the initial neural network model with adjusted weights as the first neural network model if the third work order warning level meets the work order warning information;

[0165] A training unit is used to retrain the initial neural network with adjusted weights if the third work order warning level does not conform to the work order warning information.

[0166] The embodiment of the present invention further provides an electronic device, such as Fig.10 As shown, it includes a processor 1001, a communication interface 1002, a memory 1003 and a communication bus 1004, wherein the processor 1001, the communication interface 1002, and the memory 1003 communicate with each other through the communication bus 1004.

[0167] Memory 1003, used for storing computer programs;

[0168] The processor 1001 is used to execute the program stored in the memory 1003, and implements the following steps:

[0169] Determine the work order fault content, processing notes, work order processing time limit and processing link type of the target work order, wherein the work order fault content and the processing notes are text information, and the work order processing time limit and the processing link type are digital information;

[0170] Inputting the work order fault content, the processing remarks, the work order processing time limit and the processing link type into a first neural network model to obtain a first work order warning level output by the first neural network model;

[0171] If the warning level of the first work order is greater than a preset threshold, a warning is issued;

[0172] Wherein, the first neural network model is pre-trained with sample data, and the sample data includes:

[0173] The sample work order includes the work order failure content, processing notes, work order processing time limit, processing link type and work order warning information, wherein the work order warning information is used to indicate the timeout status of the sample work order.

[0174] The communication bus mentioned in the above electronic device can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. The communication bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, only one thick line is used in the figure, but it does not mean that there is only one bus or one type of bus.

[0175] The communication interface is used for communication between the above electronic device and other devices.

[0176] The memory may include a random access memory (RAM) or a non-volatile memory (NVM), such as at least one disk memory. Optionally, the memory may also be at least one storage device located away from the aforementioned processor.

[0177] The above-mentioned processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

[0178] In another embodiment of the present invention, a computer-readable storage medium is provided, in which a computer program is stored. When the computer program is executed by a processor, the steps of any of the above-mentioned work order warning methods are implemented.

[0179] In another embodiment provided by the present invention, a computer program product including instructions is also provided, which, when executed on a computer, enables the computer to execute any work order warning method in the above embodiments.

[0180] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented by software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function described in the embodiment of the present invention is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from a website site, computer, server or data center to another website site, computer, server or data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more available media integrated. The available medium can be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid-state hard disk Solid State Disk (SSD)), etc.

[0181] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0182] Each embodiment in this specification is described in a related manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0183] The above description is only a preferred embodiment of the present invention and is not intended to limit the protection scope of the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention are included in the protection scope of the present invention.

Claims

1. A work order early warning method, characterized in that: The method comprises: Determine the work order fault content, processing notes, work order processing time limit and processing link type of the target work order, wherein the work order fault content and the processing notes are text information, and the work order processing time limit and the processing link type are digital information; Input the work order fault content, the processing remarks, the work order processing time limit and the processing link type into the first neural network model to obtain the first work order warning level output by the first neural network model, wherein the first neural network model includes: a bidirectional recurrent neural network, a unidirectional recurrent neural network, and a deep neural network, wherein the bidirectional recurrent neural network and the unidirectional recurrent neural network are used to process text information in the input of the first neural network model, and the deep neural network is used to process digital information in the input of the first neural network model; If the warning level of the first work order is greater than a preset threshold, a warning is issued; Wherein, the first neural network model is pre-trained with sample data, and the sample data includes: work order fault content, processing notes, work order processing time limit, processing link type and work order warning information of the sample work order, wherein the work order warning information is used to indicate the timeout status of the sample work order; the first neural network model is pre-trained in the following manner: setting a first learnable weight for the work order processing time limit of the sample work order, and obtaining a second learnable weight based on the first learnable weight; inputting the work order fault content, processing notes, work order processing time limit, processing link type and work order warning information of the sample work order into the initial neural network model, so that the initial neural network model outputs a second work order warning degree; adjusting the first neural network model based on the second work order warning degree and the work order warning information The first neural network model is obtained by adjusting the learnable weight, the second learnable weight, the third learnable weight, the fourth learnable weight and the fifth learnable weight; the obtaining of the first neural network model includes: inputting the work order fault content, processing remarks, work order processing time limit, processing link type and work order warning information of the verification work order into the initial neural network model with adjusted weights, so that the initial neural network model with adjusted weights outputs a third work order warning degree; verifying the third work order warning degree according to the work order warning information; if the third work order warning degree is consistent with the work order warning information, using the initial neural network model with adjusted weights as the first neural network model; if the third work order warning degree is inconsistent with the work order warning information, retraining the initial neural network with adjusted weights.

2. The method according to claim 1, characterized in that The step of inputting the work order fault content, the processing remarks, the work order processing time limit and the processing link type into a first neural network model to obtain a first work order warning level output by the first neural network model includes: Segmenting the work order fault content and the processing remarks to obtain work order fault content segmentation and processing remarks segmentation; Input the work order fault content segmentation and the processing note description segmentation into a pre-trained word library to obtain a work order fault content index and a processing note description index; Convert the work order fault content index and the processing note description index into a work order fault content word vector and a processing note description word vector; The work order fault content word vector, processing note description word vector, work order processing time limit and link status type are input into the first neural network model to obtain the first work order warning level output by the first neural network model.

3. The method according to claim 1, characterized in that Inputting the work order fault content, the processing remarks, the work order processing time limit and the processing link type into the first neural network model to obtain the first work order warning level output by the first neural network model, including: The work order fault content, the processing notes, the work order processing time limit and the processing link type are input into the first neural network model, so that the first neural network model weightedly stacks the work order processing time limit and the link status type and the first learnable weight and the second learnable weight to obtain the third learnable weight, and obtain the first work order warning level output by the first neural network model, wherein the first learnable weight is the weight corresponding to the work order processing time limit, the second learnable weight is the weight corresponding to the link status type, the fourth learnable weight is the weight corresponding to the work order fault content, the fifth learnable weight is the weight corresponding to the processing notes, the first learnable weight is negatively correlated with the second learnable weight, and the third learnable weight, the fourth learnable weight and the fifth learnable weight are negatively correlated.

4. A work order early warning device, characterized in that: The device comprises: A data acquisition module, used to determine the work order fault content, processing notes, work order processing time limit and processing link type of the target work order, wherein the work order fault content and the processing notes are text information, and the work order processing time limit and the processing link type are digital information; A warning degree acquisition module, used for inputting the work order fault content, the processing remarks, the work order processing time limit and the processing link type into a first neural network model, and obtaining the first work order warning degree output by the first neural network model, wherein the first neural network model includes: a bidirectional recurrent neural network, a unidirectional recurrent neural network, and a deep neural network, wherein the bidirectional recurrent neural network and the unidirectional recurrent neural network are used to process text information in the input of the first neural network model, and the deep neural network is used to process digital information in the input of the first neural network model; An early warning module, configured to issue an early warning if the early warning level of the first work order is greater than a preset threshold; The first neural network model is pre-trained with sample data, and the sample data includes: work order fault content, processing notes, work order processing time limit, processing link type and work order warning information of the sample work order, wherein the work order warning information is used to indicate the timeout status of the sample work order; The device also includes: a learning weight setting module, which is used to set a first learnable weight for the work order processing time limit of the sample work order, and obtain a second learnable weight based on the first learnable weight; a second work order warning acquisition module, which is used to input the work order fault content, processing notes, work order processing time limit, processing link type and work order warning information of the sample work order into the initial neural network model, so that the initial neural network model outputs a second work order warning degree; a neural network model acquisition module, which is used to adjust the first learnable weight, the second learnable weight, the third learnable weight, the fourth learnable weight and the fifth learnable weight according to the second work order warning degree and the work order warning information to obtain the first neural network model; the neural network model acquisition module includes: a third work order warning acquisition unit, which is used to input the work order fault content, processing notes, work order processing time limit, processing link type and work order warning information of the verification work order into the initial neural network model with adjusted weights, so that the initial neural network model with adjusted weights outputs a third work order warning degree; a verification unit, which is used to verify the third work order warning degree according to the work order warning information; A neural network model acquisition unit is used to use the initial neural network model with adjusted weights as the first neural network model if the third work order warning level is consistent with the work order warning information; a training unit is used to retrain the initial neural network with adjusted weights if the third work order warning level does not meet the work order warning information.

5. The device according to claim 4, characterized in that The warning degree acquisition module includes: A word segmentation unit is used to segment the work order fault content and the processing remarks to obtain work order fault content segmentation and processing remarks segmentation; An indexing unit, used to input the work order fault content segmentation and the processing note description segmentation into a pre-trained word library to obtain a work order fault content index and a processing note description index; A word vector acquisition unit, used to convert the work order fault content index and the processing note description index into a work order fault content word vector and a processing note description word vector; The warning degree acquisition unit is used to input the work order fault content word vector, processing note description word vector, work order processing time limit and link status type into the first neural network model to obtain the first work order warning degree output by the first neural network model.

6. The device according to claim 4, characterized in that The warning degree acquisition module is specifically used to input the work order fault content, the processing notes, the work order processing time limit and the processing link type into the first neural network model, so that the first neural network model weightedly stacks the work order processing time limit and the link status type and the first learnable weight and the second learnable weight to obtain the third learnable weight, and obtain the first work order warning degree output by the first neural network model, wherein the first learnable weight is the weight corresponding to the work order processing time limit, the second learnable weight is the weight corresponding to the link status type, the fourth learnable weight is the weight corresponding to the work order fault content, the fifth learnable weight is the weight corresponding to the processing notes, the first learnable weight is negatively correlated with the second learnable weight, and the third learnable weight, the fourth learnable weight and the fifth learnable weight are negatively correlated.

7. An electronic device, characterized in that: It includes a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other through the communication bus: Memory, used to store computer programs; A processor, for implementing the steps of any one of the methods described in claims 1-3 when executing a program stored in a memory.

8. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the method described in any one of claims 1 to 3 are implemented.

Images