Historical data-based user appeal association processing method and system

Abstract

The invention belongs to the field of communication, and particularly relates to a historical data-based user appeal association processing method and system, and the method comprises the steps: constructing a hierarchical historical work order library through natural language processing and a classification algorithm, and setting in a sensitive evaluation model and an intelligent matching model; performing classification and sensitivity evaluation on the user demands at the current moment; a double-layer order sending mode is set, the appeals are distributed to corresponding approval nodes according to the appeal classification and sensitivity, and historical associated work orders are searched through an intelligent matching model; distributing the historical associated work order and the basic information of the user to an examination and approval node, and carrying out monitoring and secondary distribution on an abnormal work order; performing examination and approval reply and reply auditing by using the receipt template; according to the invention, the hierarchical historical work order library is constructed and a double-layer order sending mode is adopted, so that the work order processing efficiency and accuracy are improved, and the safety of user demand processing is enhanced at the same time.

Description

Technical Field

[0001] The present invention belongs to the field of communications, and in particular relates to a method and system for processing user demand association based on historical data. Background Art

[0002] With the rapid development of information technology and the widespread application of smart grids, the business processing methods of the power industry are also constantly innovating. In recent years, State Grid has actively promoted digital transformation to improve service quality, reduce operating costs and increase customer satisfaction. The Marketing 2.0 system built under this background has successfully achieved business switching in the power industry, marking that the power marketing business processing has entered a new stage.

[0003] For example, a patent with authorization announcement number CN104767630B discloses a dispatching method and device based on alarm association, the method comprising: receiving a first alarm message sent by a network element; judging, according to a preset alarm association rule, whether there is a second alarm message having an association relationship with the first alarm message in the received alarm message; when it is determined that there is a second alarm message having an association relationship with the first alarm message, creating an association relationship between the first alarm message and the second alarm message; when a preset dispatching condition is met, dispatching a fault work order according to the association relationship, the second alarm message and the alarm message having the association relationship created with the second alarm message.

[0004] The above existing technologies have the following problems: the existing technologies only dispatch orders based on alarm information, lack flexible and diverse dispatch modes, and are difficult to adapt to different service scenarios and service needs. To solve this problem, the present invention provides a user demand association processing method and system based on historical data. Summary of the invention

[0005] In view of the shortcomings of the prior art, the present invention proposes a method and system for user demand association processing based on historical data, including: constructing a hierarchical historical work order library through natural language processing and classification algorithms, and building in a sensitivity assessment model and an intelligent matching model; classifying and evaluating the sensitivity of user demands at the current moment; setting a two-layer dispatching mode, allocating demands to corresponding approval nodes according to demand classification and sensitivity, and finding historical related work orders through an intelligent matching model; allocating historical related work orders together with user basic information to approval nodes, and monitoring and re-dispatching abnormal work orders; and using a receipt template for approval reply and reply review. The present invention improves the efficiency and accuracy of work order processing by constructing a hierarchical historical work order library and adopting a two-layer dispatching mode, while also enhancing the security of user demand processing.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] On the one hand, the present invention provides a method for processing user demand association based on historical data, comprising:

[0008] S1: Obtain the user's historical work order data and basic information data, build a historical work order library through natural language algorithms and classification algorithms, and embed a sensitive assessment model and an intelligent matching model in the interface of the constructed work order library to obtain a hierarchical historical work order library;

[0009] S2: Obtain the user demand data at the current moment, extract keywords through the natural language algorithm built into the hierarchical historical work order database, and classify the user demands at the current moment using the built-in classification algorithm. At the same time, evaluate the classified user demands through the sensitivity evaluation model to obtain the sensitivity of the corresponding user demands;

[0010] S3: Set up a two-tier dispatching mode. According to the classification results of user demands and the sensitivity of user demands, the work orders corresponding to the user demands are assigned to the corresponding approval reply nodes through the set dispatching mode, and real-time adjustments are made according to the number of reminders for the corresponding work orders. At the same time, based on the user demand data, the intelligent matching model is used to determine whether there are historical related work orders corresponding to the user demands in the hierarchical historical work order database, and the corresponding historical related work order data is obtained;

[0011] S4: The acquired historical related work order data and the user basic information corresponding to the user's request are simultaneously assigned to the approval reply node corresponding to the user's request, and the assigned work orders are monitored for exceptions and re-assigned using the set abnormal dispatch monitoring strategy;

[0012] S5: Set up a receipt template, and use the set receipt template to approve and reply to the assigned work order based on the historical related work order data and the user basic information corresponding to the user's request, and judge and modify the result of the reply review.

[0013] Specifically, the steps for constructing the historical work order library in S1 include:

[0014] S101, perform ternary annotation on the acquired historical work order data, and input the annotated historical work order data into the relationship extraction model constructed by the pre-trained Bert model, obtain the work order keywords and attribute data, and input the acquired work order keywords and attributes into the graph database to obtain the historical work order library;

[0015] S102, using a convolutional network to build a work order classification model, and inputting the work order classification keyword data and the corresponding historical user demand data in the historical work order library into the work order classification model for training, to obtain a trained work order classification model;

[0016] S103, embedding the work order classification model into the historical work order library, classifying the work orders saved in the historical work order library in real time according to the saved keywords and attribute data, and saving the classified work orders in segments according to categories;

[0017] S104. Construct a work order dispatch indication keyword set based on the obtained work order keywords and attributes, set a double-layer dispatch mode through the work order classification results and the constructed work order dispatch indication keyword set, and save the set double-layer dispatch mode into the historical work order library under the corresponding category work order attributes.

[0018] Specifically, the steps for constructing the historical work order library in S1 also include:

[0019] S105. Obtain sensitive keywords, sensitive keyword frequencies, the number of reminders for the same work order, and the number of work orders corresponding to the same event under the corresponding category from the work order keywords and attributes saved under the corresponding category, and construct a sensitivity assessment training set;

[0020] S106. Construct a sensitive assessment model based on a comprehensive fuzzy algorithm, and embed three-level assessment indicators and corresponding score ranges in the sensitive assessment model;

[0021] S107, input the acquired sensitivity assessment training set into the sensitivity assessment model for training, and build the acquired trained sensitivity assessment model into the historical work order library, perform sensitivity assessment on the work order data stored in the shards under different categories through the trained sensitivity assessment model, and obtain the sensitivity score and corresponding sensitivity level of the corresponding work order;

[0022] S108. According to the obtained sensitivity level of the work order, the work order data stored in each shard is stored in different levels, and the corresponding scoring interval and level attribute are marked under the work order data stored in each level to obtain a hierarchical historical work order library. At the same time, according to the basic information of the user corresponding to each work order in the hierarchical historical work order library and the obtained sensitivity level of the work order, the user corresponding to each work order is sensitively marked by an automatic marking algorithm;

[0023] S109, setting an alarm threshold, and real-time counting the number of work orders corresponding to the same event in the same area at each level in the hierarchical historical work order database. When the number of work orders corresponding to the same event in the same area is greater than the alarm threshold, an alarm prompt is issued for the work orders corresponding to the number greater than the alarm threshold.

[0024] Specifically, the two-tier dispatch mode includes automatic dispatch mode and manual dispatch mode;

[0025] Work order categories include special work orders and non-special work orders;

[0026] The work order dispatch instruction keyword set includes instruction keyword set 1 and instruction keyword set 2;

[0027] All special work orders containing the indicator keyword set 1 adopt the automatic dispatch mode, and send SMS or phone calls to the corresponding approval reply node;

[0028] All non-special work orders containing the indicator keyword set 2 adopt the manual dispatch mode;

[0029] The work order sensitivity levels include primary sensitivity, secondary sensitivity and tertiary sensitivity; the approval response nodes include primary approval nodes, intermediate approval nodes and senior approval nodes;

[0030] The first-level sensitivity corresponds to the work order and the primary approval node; the second-level sensitivity corresponds to the work order and the intermediate approval node; the third-level sensitivity corresponds to the work order and the advanced approval node.

[0031] Specifically, the steps of allocating the corresponding user demands to the corresponding approval reply nodes in S3 include:

[0032] S301: When a user submits a new demand, the user's basic information data and demand keyword data are extracted from the demand submitted by the user through the relationship extraction model, and the obtained demand keyword data is input into the work order classification model to obtain the work order category corresponding to the user's demand at the current moment;

[0033] S302: According to the acquired basic information data of the user and the keyword data of the demand, the intelligent matching model is used to match the historical work orders stored in the hierarchical historical work order library under the work order category corresponding to the user's demand, and a matching score is obtained;

[0034] S303, setting a matching score threshold, obtaining the user's historical work order information corresponding to the matching score greater than the matching score threshold, and obtaining the sensitivity level of the historical work order corresponding to the user's request by grading the sensitivity of the corresponding user annotation in the historical work order library based on the obtained historical work order information;

[0035] S304, based on the acquired user demand keyword data and the corresponding keyword data in the historical associated work order information, determine whether the user demand keyword data and the corresponding keyword data in the historical user work order information contain keywords in the indicator keyword set 1 or the indicator keyword set 2 through the work order assignment indication keyword set and the search algorithm, and obtain a determination result;

[0036] S305. According to the work order category, sensitivity level and judgment result corresponding to the user's request, the work order corresponding to the current user's request, the user's basic information and the user's historical related work orders are dispatched to the corresponding approval reply node through the set double-layer dispatching mode.

[0037] Specifically, the steps for constructing an abnormal dispatch monitoring strategy include:

[0038] S401, monitor the work order dispatching process in S305 in real time through the abnormal monitoring algorithm, obtain the work order type with abnormal dispatching and the number of work orders of the corresponding type in the dispatching process, if the abnormal work order type is a expedited work order among special work orders, automatically cancel the work order with abnormal dispatching within 3 minutes, and perform manual dispatching; if the abnormal work order type is not an expedited work order, automatically cancel the work order with abnormal dispatching within 20 minutes, and perform manual dispatching;

[0039] S402: If the abnormal work order obtained is a missing work order and cannot be manually dispatched, the obtained work order is placed on the top prompt, the missing work order is supplemented, and the corresponding work order after supplementation is manually dispatched;

[0040] S403. Set the remaining processing time threshold for work orders. If the abnormal work orders obtained are work orders with a remaining processing time less than the remaining processing time threshold for work orders, repeated work orders for reminders, and work orders for power outages due to arrears, then an alarm reminder will be issued for the work orders with a remaining processing time less than the remaining processing time threshold for work orders, repeated work orders for reminders, and work orders for power outages due to arrears.

[0041] Specifically, the specific steps of S5 include:

[0042] S501. When the work order assigned in the S305 process is not a work order for power restoration due to overdue payment, a corresponding return slip suggestion is generated through a text generation algorithm according to the assigned work order, historical related work order data and basic user information corresponding to the user's request, and a return slip corresponding to the user's request is obtained according to the generated return slip suggestion and the set return slip template;

[0043] S502, setting the power restoration review conditions. When the work order assigned in the S305 process is a power restoration work order for overdue payments, the power restoration review conditions are automatically judged based on the basic electricity information in the basic information of the user. If the power restoration review conditions are met, the power restoration request is automatically reviewed, the power restoration operation is triggered, and a receipt corresponding to the user's request is generated through the S501 process;

[0044] S503: If the re-call review condition is not met, the re-call operation is not triggered, and a corresponding user request receipt is generated through the S501 process, and the unsatisfied re-call review condition is built into the receipt;

[0045] S504. Build and train an automatic audit model, and embed the built automatic audit model into the audit subsystem. Automatically input the generated receipt into the audit subsystem for real-time audit. If the audit is passed, the receipt will be transferred to the next node. If the audit is not passed, the receipt will be rejected and re-edited until it passes the audit.

[0046] On the other hand, the present invention provides a user demand association processing system based on historical data, comprising: a historical work order library module, a user demand processing module and a work order dispatching module;

[0047] The historical work order library module includes a historical work order library construction unit, a work order classification unit and a hierarchical storage unit;

[0048] A historical work order database construction unit is used to obtain the historical work order data and basic information data of the user and construct the historical work order database through natural language algorithm and classification algorithm;

[0049] The work order classification unit is used to classify the saved historical and real-time work orders according to the work order keywords and attribute data saved in the historical work order library through the work order classification model;

[0050] The hierarchical storage unit is used to evaluate the sensitivity level of historical and real-time saved work orders through a sensitivity assessment model based on the sensitive keywords of the corresponding work orders after classification, the frequency of sensitive keywords, the number of reminders for the same work order, and the number of work orders corresponding to the same event under the corresponding category, and based on the assessment results, the work orders are hierarchically stored and sensitivity labeled.

[0051] Specifically, the user demand processing module includes a demand classification unit, a related work order acquisition unit, a demand evaluation unit and a determination unit;

[0052] A demand classification evaluation unit is used to obtain the work order category corresponding to the user's demand at the current moment through the work order classification model according to the acquired demand keyword information;

[0053] The associated work order acquisition unit is used to acquire the user's historical associated work order information through an intelligent matching model based on the acquired user basic information data and demand keyword data;

[0054] The demand evaluation unit is used to obtain the user's historical related work order information by grading the sensitivity of the corresponding user's annotation in the historical work order library to obtain the sensitivity level of the historical work order corresponding to the user's demand;

[0055] A determination unit is used to determine whether the user demand keyword data and the corresponding keyword data in the historical associated work order information obtained contain keywords in the indication keyword set 1 or the indication keyword set 2 according to the obtained user demand keyword data and the corresponding keyword data in the historical associated work order information obtained, and obtain a determination result through the work order assignment indication keyword set and the search algorithm;

[0056] The work order dispatching module is used to dispatch the work order corresponding to the current user demand, user basic information and user historical related work orders to the corresponding approval reply node through the set double-layer dispatching mode according to the work order category, sensitivity level and judgment result corresponding to the obtained user demand.

[0057] Finally, the present invention also provides a computer-readable storage medium having computer instructions stored thereon, which, when the computer instructions are executed, executes a user demand association processing method based on historical data.

[0058] Compared with the prior art, the present invention has the following beneficial effects:

[0059] In view of the deficiencies in the prior art, the present invention constructs a hierarchical historical work order database and combines it with a two-layer dispatching mode. It can not only flexibly allocate work orders according to the sensitivity and category of user demands, but also effectively handle user demands under different categories. In particular, it utilizes natural language processing technology and sensitivity assessment models to accurately classify and assess the sensitivity of user demands, thereby ensuring the accurate dispatch of work orders, and searching for historical related work orders through an intelligent matching model, thereby improving the efficiency and quality of problem solving. In addition, an abnormal dispatching monitoring strategy is used to ensure the accuracy and timeliness of the work order flow process, thereby avoiding service delays or waste of resources due to erroneous dispatching. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 This is a flow chart of a method for associating user demands based on historical data according to Embodiment 1 of the present invention;

[0061] Figure 2 A flowchart for constructing a hierarchical historical work order library according to Embodiment 1 of the present invention;

[0062] Figure 3 This is a module diagram of a user demand association processing system based on historical data as described in Example 2 of the present invention. DETAILED DESCRIPTION

[0063] Example 1

[0064] See also Figure 1 , an embodiment of the present invention provides: a method for processing user demand association based on historical data, comprising the following steps:

[0065] S1: Obtain the user's historical work order data and basic information data, build a historical work order library through natural language algorithms and classification algorithms, and build a sensitive assessment model and intelligent matching model into the interface of the constructed work order library to obtain a hierarchical historical work order library; For further information, please refer to Figure 2 In this embodiment, the steps of constructing the hierarchical historical work order library in S1 include:

[0066] S101, perform ternary annotation on the acquired historical work order data, and input the annotated historical work order data into the relationship extraction model constructed by the pre-trained Bert model, obtain the work order keywords and attribute data, and input the acquired work order keywords and attributes into the graph database to obtain the historical work order library;

[0067] Furthermore, in this embodiment, the historical work order data includes user basic information and user electricity usage information, the user electricity usage information includes electricity usage records, historical payment status and equipment maintenance history, etc.; the user basic information includes the electricity user number, contact information and contact address, etc.;

[0068] S102, using a convolutional network to build a work order classification model, and inputting the work order classification keyword data and the corresponding historical user demand data in the historical work order library into the work order classification model for training, to obtain a trained work order classification model;

[0069] S103, embedding the work order classification model into the historical work order library, classifying the work orders saved in the historical work order library in real time according to the saved keywords and attribute data, and saving the classified work orders in segments according to categories;

[0070] Work order categories include special work orders and non-special work orders;

[0071] Special work orders include expedited work orders, electrical appliance damage assessment work orders, overdue power restoration work orders, password reset work orders, and special emergency work orders;

[0072] Non-special work orders include opinion work orders, service application work orders, suggestion work orders, complaint work orders, praise work orders, report work orders, appointment work orders, city company rejected work orders, overtime work orders, complaint-to-opinion work orders, repeated call work orders, unfinished work orders with no processing time limit, hot sensitive word work orders, charging pile work orders and major event work orders;

[0073] S104, constructing a work order dispatch indication keyword set according to the obtained work order keywords and attributes, setting a double-layer dispatch mode according to the work order classification result and the constructed work order dispatch indication keyword set, and storing the set double-layer dispatch mode in the corresponding category work order attribute in the historical work order library;

[0074] Furthermore, the work order dispatch indication keyword set includes indication keyword set 1 and indication keyword set 2; indication keyword set 1 includes keywords such as major events, casualties, people trapped in elevators, oxygen inhalation, and ventilators; indication keyword set 2 includes keywords such as ultra-high voltage, 1000KV, 1 million volts, handheld electricity, WeChat public account, Alipay, safe distance, radiation, and noise;

[0075] Furthermore, the dual-layer dispatching mode in this embodiment includes an automatic dispatching mode and a manual dispatching mode;

[0076] Further, the correspondence between the work order type and the dispatch mode in this embodiment includes: the work orders containing the indicator keyword set 1 in the special work order all adopt the automatic dispatch mode, and push SMS or phone calls to the corresponding approval reply node; the work orders containing the indicator keyword set 2 in the non-special work order all adopt the manual dispatch mode;

[0077] S105. Obtain sensitive keywords, sensitive keyword frequencies, the number of reminders for the same work order, and the number of work orders corresponding to the same event under the corresponding category from the work order keywords and attributes saved under the corresponding category, and construct a sensitivity assessment training set;

[0078] S106. Construct a sensitive assessment model based on a comprehensive fuzzy algorithm, and embed three-level assessment indicators and corresponding score ranges in the sensitive assessment model;

[0079] Furthermore, in this embodiment, the work order sensitivity level includes primary sensitivity, secondary sensitivity and tertiary sensitivity. The primary sensitivity, secondary sensitivity and tertiary sensitivity correspond to the three-level evaluation index one by one, and the primary sensitivity corresponds to a score interval of [0, 3], the secondary sensitivity corresponds to a score interval of (3, 6], and the tertiary sensitivity corresponds to a score interval of (6, 9];

[0080] S107, input the acquired sensitivity assessment training set into the sensitivity assessment model for training, and build the acquired trained sensitivity assessment model into the historical work order library, perform sensitivity assessment on the work order data stored in the shards under different categories through the trained sensitivity assessment model, and obtain the sensitivity score and corresponding sensitivity level of the corresponding work order;

[0081] S108. According to the obtained sensitivity level of the work order, the work order data stored in each shard is stored in different levels, and the corresponding scoring interval and level attribute are marked under the work order data stored in each level to obtain a hierarchical historical work order library. At the same time, according to the basic information of the user corresponding to each work order in the hierarchical historical work order library and the obtained sensitivity level of the work order, the user corresponding to each work order is sensitively marked by an automatic marking algorithm;

[0082] S109, setting an alarm threshold, and real-time counting the number of work orders corresponding to the same event in the same region at each level in the hierarchical historical work order database. When the number of work orders corresponding to the same event in the same region is greater than the alarm threshold, an alarm prompt is issued for the work orders corresponding to the number greater than the alarm threshold;

[0083] This process uses the ternary method to annotate historical work order data and uses the pre-trained Bert model to build a relationship extraction model, which can accurately extract work order keywords and attribute data, and then build a high-quality historical work order library; the training and built-in work order classification model enables the system to quickly and accurately classify new work orders. The sharding and preservation mechanism of work orders not only improves the efficiency of data management, but also, by setting up a two-layer dispatching mode, it can intelligently assign work orders to corresponding approval reply nodes according to the classification results and sensitivity levels of work orders, ensuring the timeliness and pertinence of work order processing; the application of the sensitivity assessment model can evaluate the sensitivity of work orders, further optimize the work order processing process, and improve processing efficiency; in addition, through the automatic labeling algorithm to label users with sensitivity, and set alarm thresholds to monitor the number of work orders, it can issue an alarm prompt in time when the number of work orders reaches a certain threshold, effectively preventing potential service risks.

[0084] S2: Obtain the user demand data at the current moment, extract keywords through the natural language algorithm built into the hierarchical historical work order database, and classify the user demands at the current moment using the built-in classification algorithm. At the same time, evaluate the classified user demands through the sensitivity evaluation model to obtain the sensitivity of the corresponding user demands;

[0085] Furthermore, the user history data in this embodiment is information actively sent by the user or collected through the public website with active permission, which will not be described in detail in this embodiment;

[0086] S3: Set up a two-tier dispatching mode. According to the classification results of user demands and the sensitivity of user demands, the work orders corresponding to the user demands are assigned to the corresponding approval reply nodes through the set dispatching mode, and real-time adjustments are made according to the number of reminders for the corresponding work orders. At the same time, based on the user demand data, the intelligent matching model is used to determine whether there are historical related work orders corresponding to the user demands in the hierarchical historical work order database, and the corresponding historical related work order data is obtained;

[0087] Furthermore, in this embodiment, the approval reply node includes a primary approval node, an intermediate approval node and a senior approval node; wherein the primary approval node corresponds to an on-duty employee-level approval personnel, the intermediate approval node corresponds to an on-duty shift leader-level approval personnel, and the senior approval node corresponds to a director-level approval personnel; and when the user demands corresponding to the special work order and the non-special work order have more than two urging work orders, the level of the approval node corresponding to the work order is raised by one level from the current approval node level, and enters a higher-level node for approval; for example, the current overdue power restoration work order is approved at the primary approval node, but after the system counts the work orders, when the overdue power restoration work order has three urging work orders, the overdue power restoration work order is assigned to the intermediate approval node for approval;

[0088] Further, in this embodiment, the correspondence between sensitivity and approval nodes includes: first-level sensitivity corresponds to work orders corresponding to primary approval nodes; second-level sensitivity corresponds to work orders corresponding to intermediate approval nodes; third-level sensitivity corresponds to work orders corresponding to advanced approval nodes;

[0089] Furthermore, in this embodiment, the step of allocating the corresponding user demand to the corresponding approval reply node in S3 includes:

[0090] S301: When a user submits a new demand, the user's basic information data and demand keyword data are extracted from the demand submitted by the user through the relationship extraction model, and the obtained demand keyword data is input into the work order classification model to obtain the work order category corresponding to the user's demand at the current moment;

[0091] S302: According to the acquired basic information data of the user and the keyword data of the demand, the intelligent matching model is used to match the historical work orders stored in the hierarchical historical work order library under the work order category corresponding to the user's demand, and a matching score is obtained;

[0092] Furthermore, the intelligent matching model in this embodiment is obtained by constructing BiLstm and attention network;

[0093] S303, setting a matching score threshold, obtaining the user's historical work order information corresponding to the matching score greater than the matching score threshold, and obtaining the sensitivity level of the historical work order corresponding to the user's request by grading the sensitivity of the corresponding user annotation in the historical work order library based on the obtained historical work order information;

[0094] S304, based on the acquired user demand keyword data and the corresponding keyword data in the historical associated work order information, determine whether the user demand keyword data and the corresponding keyword data in the historical user work order information contain keywords in the indicator keyword set 1 or the indicator keyword set 2 through the work order assignment indication keyword set and the search algorithm, and obtain a determination result;

[0095] S305. According to the work order category, sensitivity level and judgment result corresponding to the user's request, the work order corresponding to the current user's request, the user's basic information and the user's historical related work orders are dispatched to the corresponding approval reply node through the set double-layer dispatching mode.

[0096] This process can quickly and accurately understand user needs by extracting and classifying keywords from user demand data, and evaluate the sensitivity of user demands through a built-in sensitivity assessment model, ensuring the security and compliance of the processing process. The setting of the two-tier dispatching mode intelligently assigns work orders to the most suitable approval response node based on the classification results and sensitivity of user demands, achieving reasonable allocation of resources; the dynamic adjustment mechanism at the approval node level can automatically upgrade the approval level according to the change in the number of work order reminders, ensuring that important or urgent work orders can be processed in a timely and effective manner; in addition, the relevance of historical work orders can be judged through an intelligent matching model, which can provide historical experience support for current work orders and improve processing efficiency.

[0097] S4: The acquired historical related work order data and the user basic information corresponding to the user's request are simultaneously assigned to the approval reply node corresponding to the user's request, and the assigned work orders are monitored for exceptions and re-assigned using the set abnormal dispatch monitoring strategy;

[0098] Furthermore, the steps of constructing the abnormal dispatch monitoring strategy in this embodiment include:

[0099] S401, monitor the work order dispatching process in S305 in real time through the abnormal monitoring algorithm, obtain the work order type with abnormal dispatching and the number of work orders of the corresponding type in the dispatching process, if the abnormal work order type is a expedited work order among special work orders, automatically cancel the work order with abnormal dispatching within 3 minutes, and perform manual dispatching; if the abnormal work order type is not an expedited work order, automatically cancel the work order with abnormal dispatching within 20 minutes, and perform manual dispatching;

[0100] Furthermore, the anomaly monitoring algorithm in this embodiment includes a random forest algorithm and an ensemble algorithm;

[0101] S402: If the abnormal work order obtained is a missing work order and cannot be manually dispatched, the obtained work order is placed on the top prompt, the missing work order is supplemented, and the corresponding work order after supplementation is manually dispatched;

[0102] S403. Set the remaining processing time threshold for work orders. If the abnormal work orders obtained are work orders with a remaining processing time less than the remaining processing time threshold for work orders, repeated work orders for reminders, and work orders for power outages due to arrears, an alarm reminder will be issued for the work orders with a remaining processing time less than the remaining processing time threshold for work orders, repeated work orders for reminders, and work orders for power outages due to arrears.

[0103] The anomaly monitoring algorithm can monitor the work order dispatching process in real time, promptly discover and correct dispatch anomalies, and ensure that work orders can be correctly assigned to the corresponding approval reply nodes. For special work orders with reminders, the abnormal dispatch can be automatically canceled within 3 minutes and quickly transferred to manual processing, reducing the user's waiting time. For other types of abnormal work orders, manual intervention will be carried out within 20 minutes to ensure that the problem is solved in a timely manner. In addition, by setting the remaining processing time threshold for work orders, alarm reminders will be issued for work orders with short remaining processing time, repeated reminders, and power outage and restoration work orders due to arrears, which will help improve processing efficiency and prevent work order delays.

[0104] S5: Set up a receipt template, and use the set receipt template to approve and reply to the assigned work order based on the historical related work order data and the user basic information corresponding to the user's request, and judge and modify the result of the reply review.

[0105] Furthermore, the specific steps of S5 in this embodiment include:

[0106] S501. When the work order assigned in the S305 process is not a work order for power restoration due to overdue payment, a corresponding return slip suggestion is generated through a text generation algorithm according to the assigned work order, historical related work order data and basic user information corresponding to the user's request, and a return slip corresponding to the user's request is obtained according to the generated return slip suggestion and the set return slip template;

[0107] Furthermore, in this embodiment, whether the work order assigned in the S305 process is an overdue power restoration work order is determined in real time by using the configured work order classification model;

[0108] S502, setting the power restoration review conditions. When the work order assigned in the S305 process is a power restoration work order for overdue payments, the power restoration review conditions are automatically judged based on the basic electricity information in the basic information of the user. If the power restoration review conditions are met, the power restoration request is automatically reviewed, the power restoration operation is triggered, and a receipt corresponding to the user's request is generated through the S501 process;

[0109] Furthermore, in this embodiment, the power restoration review conditions include the arrears status and the power outage time status, wherein when the arrears status is settled, and the power outage time status satisfies the corresponding conditions of: power outage time < 95598 work order acceptance time (arrears power restoration registration) < power restoration time < work order reply time, it is determined that the power restoration review conditions are met, and the power restoration operation is performed;

[0110] S503: If the re-call review condition is not met, the re-call operation is not triggered, and a corresponding user request receipt is generated through the S501 process, and the unsatisfied re-call review condition is built into the receipt;

[0111] S504. Build and train an automatic audit model, and embed the built automatic audit model into the audit subsystem. Automatically input the generated receipt into the audit subsystem for real-time audit. If the audit is passed, the receipt will be transferred to the next node. If the audit is not passed, the receipt will be rejected and re-edited until it passes the audit.

[0112] Furthermore, the automatic audit model in this embodiment includes an audit sub-model 1 and an audit sub-model 2, wherein the audit sub-model 1 is constructed by a random forest algorithm, and the audit sub-model 2 is constructed by pre-trained RoBERTa, and is trained using special and non-special reply work orders that have been manually corrected;

[0113] Among them, the specific audit process of the automatic audit model is as follows:

[0114] Obtain historical reply tickets, and build an entity recognition extraction model through the Chinese pre-trained Bert model. Input the historical reply tickets into the entity recognition extraction model to extract industry-related vocabulary, and use the extracted industry-related vocabulary to build an industry-related vocabulary library, and then build the constructed industry-related vocabulary library into the audit subsystem;

[0115] The work order for real-time approval is input into the review subsystem, and the type of the input review reply work order is determined through the classification model. If it is a reply work order corresponding to a special work order, the industry-related words in the work order are extracted through the entity recognition extraction model, and the extracted industry-related words are input into the review sub-model 1 for matching and comparison with the industry-related word library, and the matching result 1 is output. At the same time, the non-industry-related words are input into the review sub-model 2 for matching and comparison with the corresponding user information in the system, and the matching result 2 is output;

[0116] When the matching probabilities corresponding to matching result 1 and matching result 2 are both 1, the matching review is passed. Otherwise, the inconsistent location information in the corresponding reply work order is marked in red, and the work order is marked as abnormal and rejected;

[0117] If it is a reply work order corresponding to a non-special work order, the corresponding reply work order will be directly input into the review sub-model 2 for matching review. If the matching probability is 1, the matching review will pass. Otherwise, the inconsistent location information in the corresponding reply work order will be marked in red, and the work order will be marked as abnormal and rejected;

[0118] For example, if the current user's corresponding work order is a special work order for power restoration due to arrears, the content of the corresponding reply work order is "Reply order content: User Zhang San (user number: 123456789), power outage due to arrears, the arrears have been settled, the power outage time is 2023-03-01, apply for power restoration."

[0119] The industry-related words "power outage", "arrears", "settlement of fees", and "power restoration" are extracted through the entity recognition extraction model. The industry-related words are compared with the industry-related word library through the review sub-model 1 to see if they are consistent. At the same time, the remaining non-industry-related words, such as "user Zhang San (user number: 123456789), phone number: 151****7891", are input into the review sub-model 2 for matching and comparison with the user information in the system. If the matching probability of industry-related words and non-industry-related words is 1, the current reply work order is passed. Otherwise, the inconsistent location information in the corresponding reply work order is marked in red, and the work order is marked as abnormal and rejected; for example, if the phone number corresponding to the current user is inconsistent with that in the system, the phone number column in the reply work order will be marked in red and then rejected.

[0120] This process achieves efficient and automatic review of special and non-special work orders by introducing review sub-model 1 constructed by the random forest algorithm and review sub-model 2 constructed by the pre-trained RoBERTa model. Among them, by extracting industry-related vocabulary from historical reply work orders and building a vocabulary library, the system can accurately identify and match industry terms to ensure the correctness of professional terms. At the same time, the RoBERTa model is used to review non-industry-related vocabulary to ensure the consistency of personal information. This dual review mechanism not only improves the accuracy and efficiency of the review, but also reduces the workload of manual review, avoids the rejection of work orders due to inconsistent information, and thus improves the reliability of overall work order processing and user experience.

[0121] Example 2

[0122] See also Figure 3 , another embodiment provided by the present invention: a user demand association processing system based on historical data, comprising: a historical work order library module, a user demand processing module, a work order dispatching module, an abnormality monitoring module and a return order review module;

[0123] A historical work order library module is used to construct a hierarchical historical work order library; the historical work order library module includes a historical work order library construction unit, a work order classification unit and a hierarchical storage unit;

[0124] The historical work order library construction unit is used to construct a historical work order library through a natural language algorithm and a classification algorithm based on the historical work order data and basic information data of the user;

[0125] The work order classification unit is used to classify the saved historical and real-time work orders according to the work order keywords and attribute data saved in the historical work order library through the work order classification model;

[0126] A hierarchical storage unit is used to evaluate the sensitivity level of historical and real-time saved work orders through a sensitivity evaluation model according to the sensitive keywords of the work orders after classification, the frequency of sensitive keywords, the number of reminders for the same work order, and the number of work orders corresponding to the same event under the corresponding category, and to perform hierarchical storage and sensitivity labeling of the work orders based on the evaluation results;

[0127] The user demand processing module is used to classify and evaluate the demands submitted by users in real time, and obtain the corresponding work order categories and historical associations; the user demand processing module includes a demand classification unit, a related work order acquisition unit, a demand evaluation unit and a discrimination unit;

[0128] A demand classification evaluation unit is used to obtain the work order category corresponding to the user's demand at the current moment through the work order classification model according to the acquired demand keyword information;

[0129] The associated work order acquisition unit is used to acquire the user's historical associated work order information through an intelligent matching model based on the acquired user basic information data and demand keyword data;

[0130] The demand evaluation unit is used to obtain the user's historical related work order information by grading the sensitivity of the corresponding user's annotation in the historical work order library to obtain the sensitivity level of the historical work order corresponding to the user's demand;

[0131] A determination unit is used to determine whether the user demand keyword data and the corresponding keyword data in the historical associated work order information obtained contain keywords in the indication keyword set 1 or the indication keyword set 2 according to the obtained user demand keyword data and the corresponding keyword data in the historical associated work order information obtained, and obtain a determination result through the work order assignment indication keyword set and the search algorithm;

[0132] The work order dispatching module is used to dispatch the work order corresponding to the current user's request, user basic information and user's historical related work order to the corresponding approval reply node through the set double-layer dispatching mode according to the work order category, sensitivity level and judgment result corresponding to the user's request;

[0133] The exception monitoring module is used to detect the exceptions of the work orders assigned by the work order dispatching module in real time, and handle and re-dispatch the exceptions through the exception dispatching monitoring strategy according to the monitoring results;

[0134] The receipt review module is used to use the set receipt template to approve and reply to the assigned work orders according to the obtained historical related work order data and the basic user information corresponding to the user's request, and to judge and modify the results of the reply review. The receipt review module includes an approval unit and a review unit;

[0135] The approval unit is used to generate corresponding receipt suggestions based on the dispatched work orders, historical related work order data and user basic information corresponding to the user's request through a text generation algorithm, and obtain the receipt corresponding to the user's request based on the generated receipt suggestions and the set receipt template;

[0136] The audit unit is used to build and train the automatic audit model, and automatically input the generated receipt into the automatic audit model for real-time audit. If the audit is passed, the receipt will be transferred to the next node. If the audit is not passed, the receipt will be rejected and re-edited until it passes the audit.

[0137] Example 3

[0138] A computer-readable storage medium stores computer instructions, which, when executed, execute a user demand association processing method based on historical data.

[0139] An electronic device includes a memory and a processor, wherein the memory stores a computer program, and the processor implements a user demand association processing method based on historical data when executing the computer program.

[0140] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementation modes, which are merely illustrative rather than restrictive. Under the guidance of the present invention, ordinary technicians in the field may also change, modify, replace and modify the above-mentioned embodiments without departing from the scope of protection of the purpose of the present invention and the claims, and all of these are within the protection of the present invention.

[0141] If the disclosed technical solution involves personal information, the product using the disclosed technical solution has clearly informed the personal information processing rules and obtained the individual's voluntary consent before processing the personal information. If the disclosed technical solution involves sensitive personal information, the product using the disclosed technical solution has obtained the individual's separate consent before processing the sensitive personal information, and at the same time meets the "explicit consent" requirement. For example, on personal information collection devices such as cameras, clear and prominent signs are set to inform that the personal information collection scope has been entered and personal information will be collected. If the individual voluntarily enters the collection scope, it is deemed that he or she agrees to the collection of his or her personal information; or on the device that processes personal information, when the personal information processing rules are notified by obvious signs / information, the individual's authorization is obtained through pop-up information or by asking the individual to upload his or her personal information; among them, the personal information processing rules may include information such as the personal information processor, the purpose of personal information processing, the processing method, and the type of personal information processed.

Claims

1. A user demand association processing method based on historical data, characterized in that: include: S1: Obtain the user's historical work order data and basic information data, build a historical work order library through natural language algorithms and classification algorithms, and embed a sensitive assessment model and an intelligent matching model in the interface of the constructed work order library to obtain a hierarchical historical work order library; S2: Obtain the user demand data at the current moment, extract keywords through the natural language algorithm built into the hierarchical historical work order database, and classify the user demands at the current moment using the built-in classification algorithm. At the same time, evaluate the classified user demands through the sensitivity evaluation model to obtain the sensitivity of the corresponding user demands; S3: Set up a two-tier dispatching mode. According to the classification results of user demands and the sensitivity of user demands, the work orders corresponding to the user demands are assigned to the corresponding approval reply nodes through the set dispatching mode, and real-time adjustments are made according to the number of reminders for the corresponding work orders. At the same time, based on the user demand data, the intelligent matching model is used to determine whether there are historical related work orders corresponding to the user demands in the hierarchical historical work order database, and the corresponding historical related work order data is obtained; S4: The acquired historical related work order data and the user basic information corresponding to the user's request are simultaneously assigned to the approval reply node corresponding to the user's request, and the assigned work orders are monitored for exceptions and re-assigned using the set abnormal dispatch monitoring strategy; S5: Set up a receipt template, and use the set receipt template to approve and reply to the assigned work order based on the historical related work order data and the user basic information corresponding to the user's request, and judge and modify the result of the reply review.

2. A user demand association processing method based on historical data as claimed in claim 1, characterized in that: The steps for constructing the historical work order library in S1 include: S101, perform ternary annotation on the acquired historical work order data, and input the annotated historical work order data into the relationship extraction model constructed by the pre-trained Bert model, obtain the work order keywords and attribute data, and input the acquired work order keywords and attributes into the graph database to obtain the historical work order library; S102, using a convolutional network to build a work order classification model, and inputting the work order classification keyword data and the corresponding historical user demand data in the historical work order library into the work order classification model for training, to obtain a trained work order classification model; S103, embedding the work order classification model into the historical work order library, classifying the work orders saved in the historical work order library in real time according to the saved keywords and attribute data, and saving the classified work orders in segments according to categories; S104. Construct a work order dispatch indication keyword set based on the obtained work order keywords and attributes, set a double-layer dispatch mode through the work order classification results and the constructed work order dispatch indication keyword set, and save the set double-layer dispatch mode into the historical work order library under the corresponding category work order attributes.

3. A user demand association processing method based on historical data as claimed in claim 2, characterized in that: The steps of constructing the historical work order library in S1 also include: S105. Obtain sensitive keywords, sensitive keyword frequencies, the number of reminders for the same work order, and the number of work orders corresponding to the same event under the corresponding category from the work order keywords and attributes saved under the corresponding category, and construct a sensitivity assessment training set; S106. Construct a sensitive assessment model based on a comprehensive fuzzy algorithm, and embed three-level assessment indicators and corresponding score ranges in the sensitive assessment model; S107, input the acquired sensitivity assessment training set into the sensitivity assessment model for training, and build the acquired trained sensitivity assessment model into the historical work order library, perform sensitivity assessment on the work order data stored in the shards under different categories through the trained sensitivity assessment model, and obtain the sensitivity score and corresponding sensitivity level of the corresponding work order; S108. According to the obtained sensitivity level of the work order, the work order data stored in each shard is stored in different levels, and the corresponding scoring interval and level attribute are marked under the work order data stored in each level to obtain a hierarchical historical work order library. At the same time, according to the basic information of the user corresponding to each work order in the hierarchical historical work order library and the obtained sensitivity level of the work order, the user corresponding to each work order is sensitively marked by an automatic marking algorithm; S109, setting an alarm threshold, and real-time counting the number of work orders corresponding to the same event in the same area at each level in the hierarchical historical work order database. When the number of work orders corresponding to the same event in the same area is greater than the alarm threshold, an alarm prompt is issued for the work orders corresponding to the number greater than the alarm threshold.

4. A user demand association processing method based on historical data as claimed in claim 3, characterized in that: The dual-layer dispatching mode includes an automatic dispatching mode and a manual dispatching mode; The work order categories include special work orders and non-special work orders; The work order dispatch indication keyword set includes indication keyword set 1 and indication keyword set 2; The special work orders containing the indicator keyword set 1 are all automatically dispatched, and a text message or phone call is sent to the corresponding approval reply node; The non-special work orders containing the indicator keyword set 2 all adopt the manual dispatch mode; The work order sensitivity levels include primary sensitivity, secondary sensitivity and tertiary sensitivity; the approval reply nodes include primary approval nodes, intermediate approval nodes and senior approval nodes; The first-level sensitivity corresponds to the work order corresponding to the primary approval node; the second-level sensitivity corresponds to the work order corresponding to the intermediate approval node; the third-level sensitivity corresponds to the work order corresponding to the advanced approval node.

5. A user demand association processing method based on historical data as claimed in claim 4, characterized in that: In S3, the step of allocating the corresponding user request to the corresponding approval reply node includes: S301: When a user submits a new demand, the user's basic information data and demand keyword data are extracted from the demand submitted by the user through the relationship extraction model, and the obtained demand keyword data is input into the work order classification model to obtain the work order category corresponding to the user's demand at the current moment; S302: According to the acquired basic information data of the user and the keyword data of the demand, the intelligent matching model is used to match the historical work orders stored in the hierarchical historical work order library under the work order category corresponding to the user's demand, and a matching score is obtained; S303, setting a matching score threshold, obtaining the user's historical work order information corresponding to the matching score greater than the matching score threshold, and obtaining the sensitivity level of the historical work order corresponding to the user's request by grading the sensitivity of the corresponding user annotation in the historical work order library based on the obtained historical work order information; S304, based on the acquired user demand keyword data and the corresponding keyword data in the historical associated work order information, determine whether the user demand keyword data and the corresponding keyword data in the historical user work order information contain keywords in the indicator keyword set 1 or the indicator keyword set 2 through the work order assignment indication keyword set and the search algorithm, and obtain a determination result; S305. According to the work order category, sensitivity level and judgment result corresponding to the user's request, the work order corresponding to the current user's request, the user's basic information and the user's historical related work orders are dispatched to the corresponding approval reply node through the set double-layer dispatching mode.

6. A user demand association processing method based on historical data as claimed in claim 5, characterized in that: In step S4, the steps of constructing the abnormal dispatch monitoring strategy include: S401, monitor the work order dispatching process in S305 in real time through the abnormal monitoring algorithm, obtain the work order type with abnormal dispatching and the number of work orders of the corresponding type in the dispatching process, if the abnormal work order type is a expedited work order among special work orders, automatically cancel the work order with abnormal dispatching within 3 minutes, and perform manual dispatching; if the abnormal work order type is not an expedited work order, automatically cancel the work order with abnormal dispatching within 20 minutes, and perform manual dispatching; S402: If the abnormal work order obtained is a missing work order and cannot be manually dispatched, the obtained work order is placed on the top prompt, the missing work order is supplemented, and the corresponding work order after supplementation is manually dispatched; S403. Set the remaining processing time threshold for work orders. If the abnormal work orders obtained are work orders with a remaining processing time less than the remaining processing time threshold for work orders, repeated work orders for reminders, and work orders for power outages due to arrears, then an alarm reminder will be issued for the work orders with a remaining processing time less than the remaining processing time threshold for work orders, repeated work orders for reminders, and work orders for power outages due to arrears.

7. A user demand association processing method based on historical data as claimed in claim 6, characterized in that: The specific steps of S5 include: S501. When the work order assigned in the S305 process is not a work order for power restoration due to overdue payment, a corresponding return slip suggestion is generated through a text generation algorithm according to the assigned work order, historical related work order data and basic user information corresponding to the user's request, and a return slip corresponding to the user's request is obtained according to the generated return slip suggestion and the set return slip template; S502, setting the power restoration review conditions. When the work order assigned in the S305 process is a power restoration work order for overdue payments, the power restoration review conditions are automatically judged based on the basic electricity information in the basic information of the user. If the power restoration review conditions are met, the power restoration request is automatically reviewed, the power restoration operation is triggered, and a receipt corresponding to the user's request is generated through the S501 process; S503: If the re-call review condition is not met, the re-call operation is not triggered, and a corresponding user request receipt is generated through the S501 process, and the unsatisfied re-call review condition is built into the receipt; S504. Build and train an automatic audit model, and embed the built automatic audit model into the audit subsystem. Automatically input the generated receipt into the audit subsystem for real-time audit. If the audit is passed, the receipt will be transferred to the next node. If the audit is not passed, the receipt will be rejected and re-edited until it passes the audit.

8. A user demand association processing system based on historical data, which is used to implement a user demand association processing method based on historical data according to any one of claims 1 to 7, characterized in that: include: Historical work order library module, user demand processing module and work order dispatching module; The historical work order library module includes a historical work order library construction unit, a work order classification unit and a hierarchical storage unit; The historical work order library construction unit is used to construct a historical work order library through a natural language algorithm and a classification algorithm based on the historical work order data and basic information data of the user; The work order classification unit is used to classify the stored historical and real-time work orders according to the work order keywords and attribute data stored in the historical work order library through the work order classification model; The hierarchical storage unit is used to evaluate the sensitivity level of historical and real-time saved work orders through a sensitivity assessment model based on the sensitive keywords of the corresponding work orders after classification, the frequency of sensitive keywords, the number of reminders for the same work order, and the number of work orders corresponding to the same event under the corresponding category, and to perform hierarchical storage and sensitivity labeling of the work orders based on the assessment results.

9. A user demand association processing system based on historical data as claimed in claim 8, characterized in that: The user demand processing module includes a demand classification unit, a related work order acquisition unit, a demand evaluation unit and a determination unit; The demand classification evaluation unit is used to obtain the work order category corresponding to the user demand at the current moment through the work order classification model according to the acquired demand keyword information; The associated work order acquisition unit is used to acquire the user's historical associated work order information through an intelligent matching model based on the acquired user basic information data and demand keyword data; The demand evaluation unit is used to obtain the user's historical work order information by grading the sensitivity of the corresponding user's annotation in the historical work order library, and obtain the sensitivity level of the historical work order corresponding to the user's demand; The determination unit is used to determine whether the user demand keyword data and the corresponding keyword data in the historical associated work order information obtained contain keywords in the indication keyword set 1 or the indication keyword set 2 according to the obtained user demand keyword data and the corresponding keyword data in the historical associated work order information obtained, and obtain a determination result through the work order assignment indication keyword set and the search algorithm; The work order dispatching module is used to dispatch the work order corresponding to the current user demand, user basic information and user historical related work orders to the corresponding approval reply node through the set double-layer dispatching mode according to the work order category, sensitivity level and judgment result corresponding to the acquired user demand.

10. A computer-readable storage medium, characterized in that: Computer instructions are stored thereon, and when the computer instructions are executed, a user demand association processing method based on historical data as described in any one of claims 1-7 is executed.

Images