CN111178647B - Method, system and computer storage medium for pushing work order - Google Patents

Method, system and computer storage medium for pushing work order Download PDF

Info

Publication number
CN111178647B
CN111178647B CN201811329290.9A CN201811329290A CN111178647B CN 111178647 B CN111178647 B CN 111178647B CN 201811329290 A CN201811329290 A CN 201811329290A CN 111178647 B CN111178647 B CN 111178647B
Authority
CN
China
Prior art keywords
work order
worksheets
new work
new
pushing
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN201811329290.9A
Other languages
Chinese (zh)
Other versions
CN111178647A (en
Inventor
单彦会
刘辉
曹轲
荣玉军
罗红
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
China Mobile Communications Group Co Ltd
China Mobile Hangzhou Information Technology Co Ltd
Original Assignee
China Mobile Communications Group Co Ltd
China Mobile Hangzhou Information Technology Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by China Mobile Communications Group Co Ltd, China Mobile Hangzhou Information Technology Co Ltd filed Critical China Mobile Communications Group Co Ltd
Priority to CN201811329290.9A priority Critical patent/CN111178647B/en
Publication of CN111178647A publication Critical patent/CN111178647A/en
Application granted granted Critical
Publication of CN111178647B publication Critical patent/CN111178647B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
    • G06Q10/063—Operations research, analysis or management
    • G06Q10/0631—Resource planning, allocation, distributing or scheduling for enterprises or organisations
    • G06Q10/06311—Scheduling, planning or task assignment for a person or group
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
    • Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
    • Y02P90/30—Computing systems specially adapted for manufacturing

Abstract

The invention discloses a method, a system and a computer storage medium for pushing a work order, which are used for solving the technical problems in the prior art that the pushing is not timely and accurate enough when a service person goes to a gate to push the work order. Comprising the following steps: screening out regional gate service personnel in the region from all gate service personnel according to the region to which the new work order belongs; the method comprises the steps that a new work order is respectively built with historical work orders of each gate-up attendant in regional gate-up attendant according to time sequence, and a new work order sequence of each gate-up attendant is obtained; the historical worksheets are specified number of worksheets completed by each gate service personnel before the corresponding order receiving time of the new worksheets; based on each new work order sequence, calculating the new work order robbery probability of each gate-on service personnel for robbing the new work orders by using a trained deep learning model; screening the pushing staff based on all new work order robbing probabilities and specified thresholds corresponding to the regional gate-on service staff; and pushing the new work order to the order pushing personnel.

Description

Method, system and computer storage medium for pushing work order
Technical Field
The invention relates to the field of Internet, in particular to a method, a system and a computer storage medium for pushing a work order.
Background
With the rapid development of the mobile internet, the service industry based on location information has revolutionized, and these services are gradually becoming indispensable in our daily life, such as taxi taking, takeaway, etc.
Under the influence of the mobile internet, more and more industries begin to reconsider own business modes and operation modes by using the thinking of the mobile internet. For example, the fields of home broadband installation, intelligent networking and the like are gradually changed into the situation that users order on the app side, and after the qualified maintenance personnel rob the order, the service of going to the gate is reserved.
Work order pushing systems of a plurality of service platforms are used for referencing the technical schemes of the service platforms such as taxi taking, take-out and the like to different degrees, namely based on the current real-time position information of service personnel. For example, in industries needing to perform on-the-door service such as home broadband installation, the dependence degree of real-time position information of service personnel is far lower than that of the taxi taking and takeaway industries, but when a platform pushes a work order for service personnel, the work order is pushed based on the current real-time position of the service personnel, so that the work order of a working area familiar to the service personnel is frequently not pushed, and when the service personnel passes through some unfamiliar areas, the work order is pushed in the area, interference information is formed, so that the work order pushing of the platform is not accurate enough, and the order robbing efficiency is reduced.
At present, when some platforms solve the problem that pushing a work order based on a real-time position is not accurate enough, a function of increasing the work order is used, namely, the function of manually triggering the work order by platform service personnel, and the platform service personnel pushes the work order to be robbed in a certain range of the current position for the service personnel according to the current real-time position of the service personnel. The single pulling function can make up for the defect of the pushing function to a certain extent, but the single pulling function is manually triggered by platform service personnel, and the pulling Shan Shishi performance is poor. Therefore, the time for pushing the ticket to the upward servicer is not the same, and the user experience is reduced.
In view of this, how to timely and accurately push a list to a service person is a technical problem to be solved urgently.
Disclosure of Invention
The invention provides a method, a system and a computer storage medium for pushing a work order, which are used for solving the technical problems in the prior art that the pushing is not timely and accurate when the work order is pushed by a service person.
In order to solve the above technical problems, a technical solution of a method for pushing a work order according to an embodiment of the present invention is as follows:
screening out regional gate service personnel in the region from all gate service personnel according to the region to which the new work order belongs;
The new worksheets are respectively built with the historical worksheets of each gate-on attendant in the regional gate-on attendant according to the time sequence, and a new worksheet sequence of each gate-on attendant is obtained; the historical worksheets are the specified number of worksheets completed by the corresponding worksheets receiving time of the new worksheets;
based on each new work order sequence, calculating the new work order robbery probability of each gate-on service personnel for robbing the new work orders by using a trained deep learning model; the trained deep learning model is determined based on historical worksheets of all the service personnel;
screening the pushing staff based on all new work order robbing probabilities and specified thresholds corresponding to the area boarding staff; and pushing the new work order to the order pushing personnel.
Selecting regional door service personnel in the region from all door service personnel according to the region to which the new work order belongs; the new worksheets are respectively built with the historical worksheets of each gate-on attendant in the regional gate-on attendant according to the time sequence, so that a new worksheet sequence of each gate-on attendant is obtained; then, calculating each new work order sequence by using the trained deep learning model to obtain the new work order robbery probability of each gate-on service personnel for robbing the new work orders; and then screening the ticket pushing personnel from the regional gate service personnel based on all new work ticket robbing probabilities and specified thresholds corresponding to the regional gate service personnel, and pushing the new work ticket to the ticket pushing personnel. Therefore, a proper on-door service person (namely a pushing person) can be quickly found, and a new work order is pushed to the proper on-door service person in real time, so that the proper on-door service person can timely receive the push message of the new work order, and the corresponding efficiency of the new work order is improved.
Optionally, before screening out the area boarding service personnel in the area from all boarding service personnel according to the area to which the new work order belongs, the method further includes:
receiving order information of a user;
extracting basic information from the order information, and adding corresponding service information into the basic information to obtain the new work order; the basic information is contact information and address information of the user.
Optionally, the new worksheet is respectively assembled with the historical worksheets of each gate-up attendant in the regional gate-up attendant in a time sequence to obtain a new worksheet sequence of each gate-up attendant, which includes:
preprocessing the new work order into data which can be identified by the trained deep learning model according to a specified rule, and obtaining a preprocessed new work order;
the new work order sequence of each door operator is constructed in the following way:
starting to forward from the order receiving time corresponding to the new work order, and taking the preprocessed historical work orders corresponding to the appointed number of historical work orders of each gate-on service personnel from a database; the data stored in the database are the data of all the historical worksheets after being processed and the relation between the historical worksheets and service personnel;
The preprocessed new worksheets and the appointed number of processed historical worksheets are assembled into a sequence according to the time sequence of completion, and a new worksheet sequence of each gate-up attendant is obtained; wherein the processed new work order is the last element in the new work order sequence.
The time-based method has the advantages that the time-based method combines the worksheets (selected historical worksheets and new worksheets) of all the on-boarding service personnel into the new worksheet sequence according to the time sequence, so that the time relevance of the worksheets can be increased, the trained deep learning model is utilized to excavate the characteristics of different on-boarding service personnel, and finally, the new worksheet probability of each on-boarding service personnel is reflected, so that the accuracy of pushing the worksheets can be effectively improved.
Optionally, preprocessing the new work order into data that can be identified by the trained deep learning model according to a specified rule, and obtaining a preprocessed new work order, including:
extracting feature data from the new work order to obtain a feature data set of the new work order; the type of the characteristic data is the type corresponding to the characteristic data used by each historical work order when the trained deep learning model is trained;
Carrying out normalization processing on each characteristic data in the characteristic data set of the new work order to obtain a characteristic data set normalized by the new work order; the parameters adopted when the feature data are normalized are parameters adopted by the trained deep learning model corresponding to the types of the feature data during training;
and taking the characteristic data set normalized by the new work order as the preprocessed new work order.
Optionally, before calculating the new work order robbery probability of each gate-on service personnel to rob the new work order by using the trained deep learning model, the method further includes:
training the deep learning model in the following manner to obtain the trained deep learning model:
preprocessing each historical work order into data which can be recognized by a deep learning model according to the appointed rule, and obtaining each preprocessed historical work order;
selecting a preset number of preprocessed historical worksheets for each gate-up service personnel, and constructing a historical worksheet sequence according to the time sequence in which the worksheets can be completed;
and inputting the historical work order sequences of all the service personnel to the deep learning model for repeated iterative learning until the training error of the deep learning model is smaller than a set threshold value, and obtaining the trained deep learning model.
Optionally, screening the pushing personnel based on all new work order robbing probabilities and specified thresholds includes:
calculating the specified threshold value by using a trained threshold value screening model based on a historical work order;
screening the new work order robbing probability of which the numerical value is not smaller than the specified threshold value from the new work order robbing probability to obtain screened new work order robbing probability;
and taking the service personnel on the gate corresponding to the screened new work order robbing probability as the order pushing personnel.
When the order pushing personnel are screened, the threshold is designated to screen the service personnel on the gate in the area where the new work order belongs, so that the situation that the improper service personnel on the gate push the work order can be effectively avoided, and the generation of interference information is reduced.
Optionally, calculating the specified threshold with a trained threshold screening model includes:
extracting relevant information of each corresponding pushing person, each robbing person and acquiring person from the corresponding circulation information of each historical work order;
calculating the probability of each ticket pushing person, each ticket robbing person and the corresponding ticket robbing probability of the person corresponding to each historical work ticket through the trained deep learning model;
drawing all pushing worksheets according to all the worksheets pushing personnel, the worksheets grabbing personnel and the worksheets grabbing probabilities corresponding to the worksheets obtaining personnel, and obtaining worksheets grabbing probability distribution histograms corresponding to the worksheets;
Selecting the probability corresponding to the highest histogram in the histograms corresponding to each type of characteristic data from the robbery probability histograms of all pushed worksheets and the robbery probability histograms of all obtained worksheets respectively to obtain a first threshold value of the pushed worksheets and a second threshold value of the obtained worksheets;
and selecting a minimum value from the first threshold value and the second threshold value as the specified threshold value.
Optionally, calculating the specified threshold with a trained threshold screening model includes:
counting the number of pushing personnel pushing each historical work order;
clustering is carried out according to each characteristic data type of the work order, the number of pushing personnel of the historical work order under each characteristic data type is determined, and the number of pushing personnel corresponding to the characteristic data type with the largest number of pushing personnel in all the characteristic data types is selected as the specified threshold.
In a second aspect, an embodiment of the present invention provides a system for pushing a work order, including:
the screening module is used for screening out regional boarding service personnel in the region from all boarding service personnel according to the region to which the new work order belongs;
the construction module is used for constructing the new work orders with the historical work orders of each gate-on attendant in the regional gate-on attendant respectively according to time sequence to obtain a new work order sequence of each gate-on attendant; the historical worksheets are the specified number of worksheets completed by the corresponding worksheets receiving time of the new worksheets;
The calculation module is used for calculating the new work order robbing probability of each gate-on service personnel for robbing the new work order by using a trained deep learning model based on each new work order sequence; the trained deep learning model is determined based on historical worksheets of all the service personnel;
the pushing module is used for screening the pushing personnel based on the probability of all new work orders corresponding to the regional gate-on service personnel and the specified threshold; and pushing the new work order to the order pushing personnel.
Optionally, the screening module is further configured to:
receiving order information of a user;
extracting basic information from the order information, and adding corresponding service information into the basic information to obtain the new work order; the basic information is contact information and address information of the user.
Optionally, the building module is specifically configured to:
preprocessing the new work order into data which can be identified by the trained deep learning model according to a specified rule, and obtaining a preprocessed new work order;
the new work order sequence of each door operator is constructed in the following way:
starting to forward from the order receiving time corresponding to the new work order, and taking the preprocessed historical work orders corresponding to the appointed number of historical work orders of each gate-on service personnel from a database; the data stored in the database are the data of all the historical worksheets after being processed and the relation between the historical worksheets and service personnel;
The preprocessed new worksheets and the appointed number of processed historical worksheets are assembled into a sequence according to the time sequence of completion, and a new worksheet sequence of each gate-up attendant is obtained; wherein the processed new work order is the last element in the new work order sequence.
Optionally, the building module is further configured to:
extracting feature data from the new work order to obtain a feature data set of the new work order; the type of the characteristic data is the type corresponding to the characteristic data used by each historical work order when the trained deep learning model is trained;
carrying out normalization processing on each characteristic data in the characteristic data set of the new work order to obtain a characteristic data set normalized by the new work order; the parameters adopted when the feature data are normalized are parameters adopted by the trained deep learning model corresponding to the types of the feature data during training;
and taking the characteristic data set normalized by the new work order as the preprocessed new work order.
Optionally, the computing module is further configured to:
training the deep learning model in the following manner to obtain the trained deep learning model:
Preprocessing each historical work order into data which can be recognized by a deep learning model according to the appointed rule, and obtaining each preprocessed historical work order;
selecting a preset number of preprocessed historical worksheets for each gate-up service personnel, and constructing a historical worksheet sequence according to the time sequence in which the worksheets can be completed;
and inputting the historical work order sequences of all the service personnel to the deep learning model for repeated iterative learning until the training error of the deep learning model is smaller than a set threshold value, and obtaining the trained deep learning model.
Optionally, the pushing module is specifically configured to:
calculating the specified threshold value by using a trained threshold value screening model based on a historical work order;
screening the new work order robbing probability of which the numerical value is not smaller than the specified threshold value from the new work order robbing probability to obtain screened new work order robbing probability;
and taking the service personnel on the gate corresponding to the screened new work order robbing probability as the order pushing personnel.
Optionally, the pushing module is further configured to:
extracting relevant information of each corresponding pushing person, each robbing person and acquiring person from the corresponding circulation information of each historical work order;
Calculating the probability of each ticket pushing person, each ticket robbing person and the corresponding ticket robbing probability of the person corresponding to each historical work ticket through the trained deep learning model;
drawing all pushing worksheets according to all the worksheets pushing personnel, the worksheets grabbing personnel and the worksheets grabbing probabilities corresponding to the worksheets obtaining personnel, and obtaining worksheets grabbing probability distribution histograms corresponding to the worksheets;
selecting the probability corresponding to the highest histogram in the histograms corresponding to each type of characteristic data from the robbery probability histograms of all pushed worksheets and the robbery probability histograms of all obtained worksheets respectively to obtain a first threshold value of the pushed worksheets and a second threshold value of the obtained worksheets;
and selecting a minimum value from the first threshold value and the second threshold value as the specified threshold value.
Optionally, the pushing module is further configured to:
counting the number of pushing personnel pushing each historical work order;
clustering is carried out according to each characteristic data type of the work order, the number of pushing personnel of the historical work order under each characteristic data type is determined, and the number of pushing personnel corresponding to the characteristic data type with the largest number of pushing personnel in all the characteristic data types is selected as the specified threshold.
In a third aspect, an embodiment of the present invention further provides a system for pushing a worksheet, including:
at least one processor, and
a memory coupled to the at least one processor;
wherein the memory stores instructions executable by the at least one processor, the at least one processor performing the method of the first aspect described above by executing the instructions stored by the memory.
In a fourth aspect, embodiments of the present invention also provide a computer-readable storage medium, including:
the computer readable storage medium stores computer instructions which, when run on a computer, cause the computer to perform the method as described in the first aspect above.
Through the technical scheme in the one or more embodiments of the present invention, the embodiments of the present invention have at least the following technical effects:
in the embodiment provided by the invention, the regional gate service personnel in the region are screened out from all gate service personnel according to the region to which the new work order belongs; the new worksheets are respectively built with the historical worksheets of each gate-on attendant in the regional gate-on attendant according to the time sequence, so that a new worksheet sequence of each gate-on attendant is obtained; then, calculating each new work order sequence by using the trained deep learning model to obtain the new work order robbery probability of each gate-on service personnel for robbing the new work orders; and then screening the ticket pushing personnel from the regional gate service personnel based on all new work ticket robbing probabilities and specified thresholds corresponding to the regional gate service personnel, and pushing the new work ticket to the ticket pushing personnel. Therefore, a proper on-door service person (namely a pushing person) can be quickly found, and a new work order is pushed to the proper on-door service person in real time, so that the proper on-door service person can timely receive the push message of the new work order, and the corresponding efficiency of the new work order is improved.
Furthermore, in the embodiment provided by the invention, the training deep learning model is used, so that the new work order robbery efficiency of each boarding attendant in the area to which the new work order belongs can be rapidly calculated, and the proper boarding attendant (namely, the pushing personnel) is selected from the new work order robbery efficiency by combining the designated threshold, thereby improving the efficiency of screening and pushing personnel.
Furthermore, in the embodiment provided by the invention, the time relevance of the worksheets can be increased by combining the worksheets (the selected historical worksheets and the new worksheets) of each on-boarding attendant into the new worksheet sequence according to the time sequence, so that the characteristics of different on-boarding attendant are mined by using a trained deep learning model, and finally, the accuracy of pushing the worksheets can be effectively improved by reflecting the probability of the new worksheets of each on-boarding attendant.
Further, when the order pushing personnel are screened, the threshold is designated to screen the service personnel on the gate in the area where the new work order belongs, so that the situation that the improper service personnel on the gate push the work order can be effectively avoided, and the generation of interference information is reduced.
Drawings
FIG. 1 is a flowchart of a pushing work order according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of a new work order sequence provided in an embodiment of the present application;
FIG. 3 is a schematic diagram of determining a first threshold according to an embodiment of the present application;
FIG. 4 is a schematic diagram of determining a specified threshold according to an embodiment of the present application;
fig. 5 is a schematic structural diagram of a system for pushing a work order according to an embodiment of the present application.
Detailed Description
The embodiment of the application provides a method, a system and a computer storage medium for pushing a work order, which are used for solving the technical problems in the prior art that pushing is not timely and accurate enough when a service person goes to a gate to push the work order.
The technical scheme in the embodiment of the application aims to solve the technical problems, and the overall thought is as follows:
the method for pushing the work order comprises the following steps: screening out regional gate service personnel in the region from all gate service personnel according to the region to which the new work order belongs; the method comprises the steps that a new work order is respectively built with historical work orders of each gate-up attendant in regional gate-up attendant according to time sequence, and a new work order sequence of each gate-up attendant is obtained; the historical worksheets are specified number of worksheets completed by each gate service personnel before the corresponding order receiving time of the new worksheets; based on each new work order sequence, calculating the new work order robbery probability of each gate-on service personnel for robbing the new work orders by using a trained deep learning model; the trained deep learning model is determined based on historical worksheets of all the service personnel; screening the pushing staff based on all new work order robbing probabilities and specified thresholds corresponding to the regional gate-on service staff; and pushing the new work order to the order pushing personnel.
In the scheme, the regional door service personnel in the region are screened out from all the door service personnel according to the region to which the new work order belongs; the new worksheets are respectively built with the historical worksheets of each gate-on attendant in the regional gate-on attendant according to the time sequence, so that a new worksheet sequence of each gate-on attendant is obtained; then, calculating each new work order sequence by using the trained deep learning model to obtain the new work order robbery probability of each gate-on service personnel for robbing the new work orders; and then screening the ticket pushing personnel from the regional gate service personnel based on all new work ticket robbing probabilities and specified thresholds corresponding to the regional gate service personnel, and pushing the new work ticket to the ticket pushing personnel. Therefore, a proper on-door service person (namely a pushing person) can be quickly found, and a new work order is pushed to the proper on-door service person in real time, so that the proper on-door service person can timely receive the push message of the new work order, and the corresponding efficiency of the new work order is improved.
In order to better understand the above technical solutions, the following detailed description of the technical solutions of the present invention is made by using the accompanying drawings and specific embodiments, and it should be understood that the specific features of the embodiments and the embodiments of the present invention are detailed descriptions of the technical solutions of the present invention, and not limiting the technical solutions of the present invention, and the technical features of the embodiments and the embodiments of the present invention may be combined with each other without conflict.
Referring to fig. 1, an embodiment of the present invention provides a method for pushing a work order, and a processing procedure of the method is as follows.
Step 101: and screening out the regional gate service personnel in the region from all gate service personnel according to the region to which the new work order belongs.
The door-up service personnel can be personnel for door-up installation service, service personnel for door-up maintenance, or other personnel needing door-up service.
Before executing step 101, after the user places an order, the user needs to receive order information of the user, then extract basic information from the order information, and add corresponding service information into the basic information to obtain a new work order; the basic information is contact information and address information of the user.
For example, taking the installation of a broadband as an example, a user needs to fill in his name, phone, address, package type of installation of the broadband, etc. when placing an order. After receiving the order information of the user, basic information such as name, telephone, address, time for desiring to install and the like is extracted from the information, the type of the work order, release time, time for desiring to install are determined according to the order information, installation bandwidth is determined according to the package type of the broadband required to be installed selected by the user, longitude and latitude of the work order are determined according to the address of the user, the type of the work order, release time, time for desiring to install, installation bandwidth, longitude and latitude and the like determined in the order information are used as service information, and the service information is added to the basic information or is associated with the basic information, so that a new work order can be obtained. After the new work order is obtained, the new work order can be collected into a database for later analysis and processing, etc.
After the new work order is obtained, step 101 may be executed, i.e. the regional personnel in the region to which the new work order belongs are selected from all the personnel for service to which the new work order belongs according to the region to which the new work order belongs.
For example, still taking the previous example of installing a broadband as an example, assume that the address of the user is a XX cell, the dimension of the XX cell is determined to be (116.41, 39.82) according to the address, the cell in which the user is located is determined to be an XXX zone in a new work order according to the longitude and latitude (116.41, 39.82) in the new work order, and then the top installer (for example, zhang three, liu four, wang Mou) with the service area in the XXX zone is selected as the top installer of the area of the new order.
After screening out the regional gate service personnel of the region to which the new work order belongs, the gate service personnel to which the new work order is pushed need to be selected from the regional gate service personnel, and the method can be determined in steps 102-104.
Step 102: the method comprises the steps that a new work order is respectively built with historical work orders of each gate-up attendant in regional gate-up attendant according to time sequence, and a new work order sequence of each gate-up attendant is obtained; the historical worksheets are specified number of worksheets completed by each gate service personnel before the corresponding order receiving time of the new worksheets.
Specifically, the new work order can be preprocessed into data which can be recognized by a trained deep learning model according to a specified rule, and the preprocessed new work order can be obtained.
The method comprises the steps of preprocessing a new work order, extracting characteristic data from the new work order, and obtaining a characteristic data set of the new work order; the type of the characteristic data is the type corresponding to the characteristic data used by each historical work order when the trained deep learning model is trained; carrying out normalization processing on each characteristic data in the characteristic data set of the new work order to obtain a characteristic data set normalized by the new work order; the parameters adopted when each feature data is normalized are parameters adopted by the type of the corresponding feature data when the training is carried out, wherein the parameters are trained deep learning models; and finally, taking the characteristic data set normalized by the new work order as the preprocessed new work order.
For example, in a new work order, a name is recorded: zhang XX, number of mobile phone: 123456789 longitude and latitude: (116.41, 39.82), worksheet type: installation class, bandwidth: since the trained deep learning model is 100MB trained by using the characteristic data of the historical worksheet, when the new worksheet is subjected to AND processing, the type of the characteristic data of the selected new worksheet should be the same as the type of the characteristic data used by the historical worksheet, and if the type of the characteristic data of the extracted historical worksheet is longitude and latitude, the type of the worksheet and bandwidth, the new worksheet should also extract characteristic data corresponding to the longitude and latitude, the type of the worksheet and the bandwidth when the new worksheet is subjected to pretreatment, namely (116.41, 39.82), the installation class and 100MB, and the characteristic data which is text information is converted into numbers, and if the number corresponding to the installation class is 1, the installation class is converted into 1.
The feature data sets of the new worksheet are (116.41, 39.82), 1, 100MB. And carrying out normalization processing on the characteristic data, so that the value range of each characteristic data is within a specific range (such as a range of 0-1), and obtaining a characteristic data set normalized by a new work order. The range value of each type of feature data used in the normalization processing is the same as the range value used in the deep training.
The normalization formula adopted is: nor_val= (origin_val-min_val)/(max_val-min_val)
Wherein nor_val is normalized data, origin_val is a value of one feature data of the new work order, max_val is a maximum value in the type of feature data where origin_val is located, and min_val is a minimum value in the type of feature data where origin_val is located. When the data of the historical work order is normalized, the adopted calculation formula is the same as the formula adopted when the new work order is normalized.
For example, taking normalization of latitude features in a new work order as an example, it is necessary to use maximum and minimum values that are used when normalizing history data before training a depth training model when normalizing latitude. Assume that the latitude characteristic features a maximum of 45.5 and a minimum of 15.8. When the latitude (39.82) in the new work order is normalized, and the maximum value 45.5 and the minimum value 15.8 are also used, the nor_val= (39.82-15.8)/(45.5-15.8) = 24.02/29.7=0.808, and normalized characteristic data corresponding to the latitude is obtained. The other calculation process of feature data normalization is similar to the above, and will not be described again. And taking the normalized values of all the characteristic data as a characteristic data set normalized by the new work order, namely the preprocessed new work order.
Then, a new work order sequence of each door operator is built in the following way:
firstly, starting from the order receiving time corresponding to the new work order, taking the preprocessed historical work orders corresponding to the appointed number of historical work orders of each on-door service personnel from a database; the data stored in the database are the data of all the historical worksheets after being processed and the relation between the historical worksheets and service personnel; then, the processed new worksheets and the appointed number of processed historical worksheets are assembled into a sequence according to the time sequence of completion, and a new worksheet sequence of each gate-up service person is obtained; the processed new work order is the last element in the new work order sequence.
The new work order sequence is one data (also called one sample) of the trained deep learning model, so that a plurality of work orders in the new work order sequence are arranged in time sequence, the work orders in the new work order sequence belong to the same service person on the gate, the number of the work orders in the new work order sequence of the same service person on the gate is fixed, for example, the new work order sequence of the service person on the gate can be composed of 10 work orders at most, if the number of the work orders completed by the service person on the gate A obtained from a database is only 5, the last unfinished work orders of the service person on the gate can be used for filling the residual positions, and when the number of the work orders in the new work order sequence is only one, the time correlation characteristic of the work orders is not needed to be considered, and the distribution of the work orders in a multidimensional space is concerned.
For example, the order receiving time of the new work order is 2018.10.29:12, it is determined that the third, fourth and Wang Mou work orders in the area are the last serviceman in the area, and the history work order completed before the third work order is 2018.10.29:12, which is a history work order a1 (2018.10.29:10:12), a history work order a2 (2018.10.28:12), a history work order a3 (2018.10.28:16:02) … …, a history work order a20 (2018.10.19:32), the history work order completed before the fourth work order is 2018.10.29:12, which is a history work order b1 (2018.10.29:22), a history work order b2 (2018.10.29:42), a history work order b3 (2018.10.28:52), a3 (2018.3521), a history work order b20 (2018.10.28:52), a history work order b20 (2018.10.29:32), and a history work order b2 (2018.10.29:12:14) which is a history work order b20 (2018.10.29:12), which is a history work order b2 (2018.10.29:12, 2018.29:12), which is completed before the fourth work order b 14:12, which is a last work order b2 (2018.10.10.29:12, 2018.29:12, b.29:12).
If the designated number is 3, taking the new work order sequence of calculating Zhang San as an example, 3 history work orders closest to the time of receiving the new work orders (all of which are the preprocessed history work orders) need to be selected from all the history work orders of Zhang San, namely, the history work orders a1, a history work order a2, a history work order a3 and the preprocessed new work orders are constructed into a sequence (called a sequence), namely, the new work order sequence of Zhang San is: the history work order a1, the history work order a2, the history work order a3, and the pre-processed new work order are shown in fig. 2, which is a schematic diagram of a new work order sequence of Zhang san (assuming that the work order number n=4 that can be accommodated in the work order sequence of Zhang san). Similarly, new work order sequences of Lifour and Wang Mou can be obtained respectively, and are not described herein.
If the deep learning model is trained, the work orders in the adopted sequence are historical work orders, and the work orders in the work order sequence are the work orders after data preprocessing is completed, namely the work orders after numerical conversion and normalization processing are completed.
After obtaining a new work order sequence for each of the gate service personnel in the area to which the new work order belongs, step 103 may be performed.
Step 103: based on each new work order sequence, calculating the new work order robbery probability of each gate-on service personnel for robbing the new work orders by using a trained deep learning model; the trained deep learning model is determined based on historical worksheets of all the service personnel.
Before using the trained deep learning model to calculate the new work order robbery probability of each service person on the new work order, training the deep learning model according to the following mode to obtain the trained deep learning model:
first, each history work order is preprocessed into data which can be recognized by a deep learning model according to a specified rule, and each preprocessed history work order is obtained.
After the historical worksheets of all the service personnel are obtained, the useful characteristic data in the historical worksheets are extracted, and the extracted historical characteristic data are converted into numerical values according to the specified rules after the extraction, so that the preprocessed historical worksheets corresponding to all the historical worksheets are obtained.
After the historical worksheets are obtained from the database, the data in the historical worksheets need to be analyzed and a plurality of characteristics are screened out, and the characteristics can be used for distinguishing the characteristics of different worksheets and are also interesting characteristics of the service personnel on the gate, and the characteristics influence whether the service personnel on the gate participate in rescuing.
For example, a user address in a work order is a very important feature, and a gate attendant does not rob a work order that is very far from itself. For another example, price information in a work order is also an important feature, and when different orders are to be taken, the operator can give priority to the work order with high price. For another example, the type of the work order FTTB, FTTH, etc. may affect how easy the work order is to be installed, which may also affect whether the service personnel is to rob the work order. Through the analysis process similar to the above, all the features in one work order can be analyzed, and those features affecting the order robbery of the service personnel on the door are screened out as feature data. And when the characteristic data is selected for the new work order, directly selecting the characteristic data of the corresponding characteristic type for the new work order according to the screening result.
After the characteristic data of each historical work order are selected, the extracted historical characteristic data can be converted into numerical values according to a specified rule, so that the preprocessed historical work order corresponding to each historical work order is obtained.
Specifically, the specified rule is specifically 1) converting a number of a character (string) type into a numerical value of a corresponding type; 2) Converting the multiple categories into several different values; 3) The categories and combinations of categories are converted to different numerical values.
For example, the bandwidth of the broadband in the worksheet may be 50M, 100M, etc., and the data stored in the database is stored in a string type, and after reading the bandwidth data from the database, the string type 50M needs to be converted into a floating point type 50.
As another example, the network installation type includes: fiber to the building (Fiber To The Building, FTTB), fiber to the home (Fiber To The Home, FTTH), fiber to the desktop (Fiber To The Desktop, FTTD) are three types, FTTB is greatly different from FTTH, and a maintenance person may have a preference for the installation type of the network when selecting a work order, and different installation types may affect the probability of the maintenance person to rob the order, so that the three text types need to be converted into numerical types. The FTTB, FTTH, FTTD can be sequentially converted into integers 1,2 and 3 by adopting the simplest conversion mode, so that the network installation type can be identified by the deep learning model.
For another example, a worksheet may be of a type having one category, two categories, or a combination of a plurality of categories. The following four cases 1 may exist for example in the business_type of a work order; 2,5;1,2,5;11, four possible values for business_type are listed, for which a work order has a combination of categories, the following conversion formula can be used to calculate the categories as a value (so that each combination can be distinguished and the newly added category is not affected):
f(x)＝x 1 ×x 2 ×…×x n +(x 1 +x 2 +…+x n )+δ
Wherein x is n kinds of x 1 、x 2 、…、x n Delta is a constant to ensure that the converted result is not identical to one of the classes.
For example, taking a combination of three classes where x is 1,2,5 as an example, δ=100 is functionally transformed, and f (1, 2, 5) =1×2×5+ (1+2+5) +100=118.
And then, selecting a preset number of preprocessed historical worksheets for each gate-up service personnel, and constructing a historical worksheet sequence according to the time sequence of worksheet completion.
And if the preset number is 10, constructing the preprocessed historical worksheets corresponding to the 10 last completed historical worksheets of each gate service personnel into a historical worksheet sequence according to the sequence order, so as to obtain the historical worksheet sequence of each gate service personnel.
And finally, inputting the historical work order sequences of all the service personnel to be on the gate into the deep learning model for repeated iterative learning until the training error of the deep learning model is smaller than a set threshold value, and obtaining a trained deep learning model.
Deep learning models are typically based on complex neural network structures to simulate the cognitive and learning capabilities of the human brain, and through learning of a large amount of historical data, the associated information present in the data is discovered, thereby giving predictions for new data. In the process of training the deep learning model, the error between the output result and the true value is calculated in each iteration, and then the error is reversely propagated to the network of the deep learning model, so that the weight parameter and the deviation parameter of the model are updated, and then the iteration is performed again. And (3) enabling the overall error of the model to be smaller than a set threshold through continuous iteration, wherein parameters obtained by training the depth model at the moment are training models obtained after the deep learning.
Preferably, the deep learning model may be a Long Short Term Memory network (LSTM).
Because the LSTM network has memory characteristics relative to other deep learning networks, the LSTM network can store the previously processed work orders, so that the current work orders are jointly processed, and the LSTM network judges whether to use the previously unprocessed work orders according to the association characteristics among the work orders. For example, when a service man needs to go to a gate for service in cell a, there is a new work order of cell a, but the current real-time position of the maintenance man is far away from cell a, if the maintenance man does not meet the pushing condition according to the pushing rule in the prior art, but according to the work order pushing system for learning by the LSM network used in the embodiment of the present application, the work order can be pushed to the service man, because the probability that the maintenance man robs the work order can be greatly increased by the work order which is not processed before.
After the deep learning model is repeatedly and iteratively trained by using the historical work order to obtain a trained deep learning model, the robbery probability of each gate-up service personnel in the area to which the new work order belongs can be calculated by using the trained deep learning model after the new work order arrives. And then, after the probability of robbing the new work order from each of the service personnel in the area to which the new work order belongs is calculated, the service personnel to which the new work order is pushed are selected, and in particular, please refer to step 104.
Step 104: screening the pushing staff based on all new work order robbing probabilities and specified thresholds corresponding to the regional gate-on service staff; and pushing the new work order to the order pushing personnel.
Specifically, a screening and pushing person can calculate a specified threshold value by using a trained threshold value screening model based on a historical work order; screening the new work order robbing probability of which the value is not smaller than a specified threshold value from the new work order robbing probability to obtain the screened new work order robbing probability; and finally, taking the service personnel on the gate corresponding to the probability of the new work order robbery after screening as the order pushing personnel.
For example, if the calculated specified threshold is 0.8, a new work order robbery probability not smaller than 0.8 is selected from the new work orders as a new work order robbery probability after screening (assuming that the new work order robbery probability of Zhang three is 0.85, the new work order robbery probability of Liu four is 0.93, and the new work order robbery probability of Wang Mou is 0.75, the new work order robbery probability after screening is 0.85 and 0.95), and the upper gate service personnel (Zhang three and Liu four) corresponding to the new work order robbery probability after screening is used as a single pushing personnel, and the new work order is pushed to Zhang three and Liu four.
Specifically, the following two methods are used to calculate the specified threshold by using the trained threshold screening model:
Firstly, extracting relevant information of each corresponding pushing person, each robbing person and acquiring person from circulation information corresponding to each historical work order; calculating the probability of each ticket pushing person, each ticket robbing person and the corresponding ticket robbing probability of the person corresponding to each historical work ticket through the trained deep learning model; drawing all pushing worksheets according to all the worksheets pushing personnel, the worksheets grabbing personnel and the worksheets grabbing probabilities corresponding to the worksheets grabbing personnel, and obtaining worksheets grabbing probability distribution histograms corresponding to the worksheets; then, selecting probabilities corresponding to the highest histogram in the histograms corresponding to each type of characteristic data from the robbery probability histograms of all pushed worksheets and the robbery probability histograms of all obtained worksheets respectively to obtain a first threshold value of the pushed worksheets and a second threshold value of the obtained worksheets; and finally, selecting the minimum value from the first threshold value and the second threshold value as a specified threshold value.
The circulation information is information that a work order is acquired, which is extracted from a log system or a log file, and includes time for pushing the work order, ID of a pushing person, ID of an entry service person participating in the order taking, ID of a person obtaining the work order finally, and the like.
For example, the latest 100 completed historical worksheets 1-100 are obtained from the database, and the corresponding circulation information (corresponding to circulation information 1-100) of each historical worksheet is obtained, for example, the corresponding circulation information 1 is extracted from the historical worksheet 1, and the historical worksheets 1 pushed to which service personnel are moved from the circulation information 1 are obtained, assuming that the history work order 1 is pushed to the sheet XX, the thank XX, the Liao XX and the Liao XX, the sheet pushing personnel extracted from the circulation information 1 corresponding to the history work order 1 are the sheet XX, the thank XX, the Liao XX and the Liao XX, the sheet grabbing personnel actually participating in the sheet grabbing of the history work order 1 are the hand k XX, the Liao XX and the Liao XX, and finally the sheet grabbing personnel to the history work order 1 are the Liu XX. And then, calculating the probability of each ticket pushing person, each ticket robbing person and the ticket robbing probability corresponding to the person corresponding to the historical work ticket 1 by using the trained deep learning model. The determination and the robbery probability of the single pushing personnel, the single robbery personnel and the single acquisition personnel of each of the historical worksheets 2-100 are similar to those of the historical worksheet 1, and are not repeated here.
Then, the ticket pushing personnel and the ticket grabbing probability corresponding to the personnel in the historical worksheet 1-100 are respectively drawn into corresponding ticket grabbing probability distribution histograms, the drawing width of the histograms can be a histogram with 5% probability, and the height is a multiple of 5%, namely the histograms are used for representing the ticket grabbing probability of various types of characteristic data in the corresponding worksheets. From the robbery probability histograms corresponding to the pushing operators, selecting the highest histogram in the histograms corresponding to each type of feature data as the threshold value of the type of feature data, and forming a first threshold value of the pushing worksheet by using the threshold value sets of different types, wherein fig. 3 is a schematic diagram for determining the first threshold value. The manner of obtaining the second threshold of the work order is similar to that of obtaining the first threshold, and will not be described in detail herein. And taking the minimum value of the first threshold value and the second threshold value as a specified threshold value of the screening order pushing personnel. The regional gate service personnel with the probability of robbing the ticket larger than the specified threshold value are determined to be the ticket pushing personnel, and a push message of a new work ticket is sent to the regional gate service personnel.
In addition to determining the specified threshold value in the manner of calculating the probability described above, the specified threshold value may be calculated in the following manner.
Secondly, counting the number of pushing personnel pushing each historical work order; clustering is carried out according to each characteristic data type of the worksheet, the number of pushing personnel of the historical worksheets under each characteristic data type is determined, and the number of pushing personnel corresponding to the characteristic data type with the largest number of pushing personnel in all the characteristic data types is selected as a specified threshold.
For example, referring to fig. 4, the most recent 100 completed history worksheets (history worksheets 1 to 100) are selected from the database, the number of pushing personnel in each history worksheet is counted, and then the number of pushing personnel under each feature type (type 1 to type 5) is determined by clustering according to the type of the feature data, and the number of pushing personnel corresponding to the type 3 of the history worksheet 1 with the largest number of pushing personnel is selected as the specified threshold. Ordering the new worksheets of the regional gate service personnel from large to small according to the probability of the new worksheets, taking the number of people with the specified threshold value with the highest ordering as the worksheet pushing personnel, and sending push messages of the new worksheets to the people.
Through using the trained threshold screening model, circulation information in the historical worksheet is learned, a designated threshold value when the worksheet is pushed is obtained, and a proper pushing person is screened out from the probability of the new worksheet robbing the worksheet by using the designated threshold value, so that the proper pushing person can be screened out quickly and timely, the new worksheet is pushed, and the response rate of the service person on the gate to the new worksheet is improved.
Based on the same inventive concept, in an embodiment of the present invention, a system for pushing a work order is provided, and a specific implementation manner of a method for pushing a work order of the system may refer to a description of an embodiment portion of the method, and details are not repeated, and please refer to fig. 5, where the system includes:
the screening module 501 is configured to screen out the area boarding attendant in the area from all boarding attendant according to the area to which the new work order belongs;
the building module 502 is configured to build the new worksheet with a historical worksheet of each of the regional on-boarding attendant respectively in time sequence, so as to obtain a new worksheet sequence of each on-boarding attendant; the historical worksheets are the specified number of worksheets completed by the corresponding worksheets receiving time of the new worksheets;
a calculating module 503, configured to calculate, based on each new work order sequence, a new work order robbery probability of each service person who goes on the gate to rob the new work order with a trained deep learning model; the trained deep learning model is determined based on historical worksheets of all the service personnel;
The pushing module 504 is configured to screen a pushing person based on all new work order robbery probabilities and specified thresholds corresponding to the area boarding attendant; and pushing the new work order to the order pushing personnel.
Optionally, the screening module 501 is further configured to:
receiving order information of a user;
extracting basic information from the order information, and adding corresponding service information into the basic information to obtain the new work order; the basic information is contact information and address information of the user.
Optionally, the building module 502 is specifically configured to:
preprocessing the new work order into data which can be identified by the trained deep learning model according to a specified rule, and obtaining a preprocessed new work order;
the new work order sequence of each door operator is constructed in the following way:
starting to forward from the order receiving time corresponding to the new work order, and taking the preprocessed historical work orders corresponding to the appointed number of historical work orders of each gate-on service personnel from a database; the data stored in the database are the data of all the historical worksheets after being processed and the relation between the historical worksheets and service personnel;
The preprocessed new worksheets and the appointed number of processed historical worksheets are assembled into a sequence according to the time sequence of completion, and a new worksheet sequence of each gate-up attendant is obtained; wherein the processed new work order is the last element in the new work order sequence.
Optionally, the building module 502 is further configured to:
extracting feature data from the new work order to obtain a feature data set of the new work order; the type of the characteristic data is the type corresponding to the characteristic data used by each historical work order when the trained deep learning model is trained;
carrying out normalization processing on each characteristic data in the characteristic data set of the new work order to obtain a characteristic data set normalized by the new work order; the parameters adopted when the feature data are normalized are parameters adopted by the trained deep learning model corresponding to the types of the feature data during training;
and taking the characteristic data set normalized by the new work order as the preprocessed new work order.
Optionally, the calculating module 503 is further configured to:
training the deep learning model in the following manner to obtain the trained deep learning model:
Preprocessing each historical work order into data which can be recognized by a deep learning model according to the appointed rule, and obtaining each preprocessed historical work order;
selecting a preset number of preprocessed historical worksheets for each gate-up service personnel, and constructing a historical worksheet sequence according to the time sequence in which the worksheets can be completed;
and inputting the historical work order sequences of all the service personnel to the deep learning model for repeated iterative learning until the training error of the deep learning model is smaller than a set threshold value, and obtaining the trained deep learning model.
Optionally, the pushing module 504 is specifically configured to:
calculating the specified threshold value by using a trained threshold value screening model based on a historical work order;
screening the new work order robbing probability of which the numerical value is not smaller than the specified threshold value from the new work order robbing probability to obtain screened new work order robbing probability;
and taking the service personnel on the gate corresponding to the screened new work order robbing probability as the order pushing personnel.
Optionally, the pushing module 504 is further configured to:
extracting relevant information of each corresponding pushing person, each robbing person and acquiring person from the corresponding circulation information of each historical work order;
Calculating the probability of each ticket pushing person, each ticket robbing person and the corresponding ticket robbing probability of the person corresponding to each historical work ticket through the trained deep learning model;
drawing all pushing worksheets according to all the worksheets pushing personnel, the worksheets grabbing personnel and the worksheets grabbing probabilities corresponding to the worksheets obtaining personnel, and obtaining worksheets grabbing probability distribution histograms corresponding to the worksheets;
selecting the probability corresponding to the highest histogram in the histograms corresponding to each type of characteristic data from the robbery probability histograms of all pushed worksheets and the robbery probability histograms of all obtained worksheets respectively to obtain a first threshold value of the pushed worksheets and a second threshold value of the obtained worksheets;
and selecting a minimum value from the first threshold value and the second threshold value as the specified threshold value.
Optionally, the pushing module 504 is further configured to:
counting the number of pushing personnel pushing each historical work order;
clustering is carried out according to each characteristic data type of the work order, the number of pushing personnel of the historical work order under each characteristic data type is determined, and the number of pushing personnel corresponding to the characteristic data type with the largest number of pushing personnel in all the characteristic data types is selected as the specified threshold.
Based on the same inventive concept, the embodiment of the invention provides a system for pushing a work order, which comprises: at least one processor, and
a memory coupled to the at least one processor;
wherein the memory stores instructions executable by the at least one processor, the at least one processor executing the method of pushing a work order as described above by executing the instructions stored by the memory.
Based on the same inventive concept, an embodiment of the present invention also provides a computer readable storage medium, including:
the computer readable storage medium stores computer instructions that, when executed on a computer, cause the computer to perform the method of pushing work orders as described above.
In the embodiment provided by the invention, the regional gate service personnel in the region are screened out from all gate service personnel according to the region to which the new work order belongs; the new worksheets are respectively built with the historical worksheets of each gate-on attendant in the regional gate-on attendant according to the time sequence, so that a new worksheet sequence of each gate-on attendant is obtained; then, calculating each new work order sequence by using the trained deep learning model to obtain the new work order robbery probability of each gate-on service personnel for robbing the new work orders; and then screening the ticket pushing personnel from the regional gate service personnel based on all new work ticket robbing probabilities and specified thresholds corresponding to the regional gate service personnel, and pushing the new work ticket to the ticket pushing personnel. Therefore, a proper on-door service person (namely a pushing person) can be quickly found, and a new work order is pushed to the proper on-door service person in real time, so that the proper on-door service person can timely receive the push message of the new work order, and the corresponding efficiency of the new work order is improved.
Furthermore, in the embodiment provided by the invention, the training deep learning model is used, so that the new work order robbery efficiency of each boarding attendant in the area to which the new work order belongs can be rapidly calculated, and the proper boarding attendant (namely, the pushing personnel) is selected from the new work order robbery efficiency by combining the designated threshold, thereby improving the efficiency of screening and pushing personnel.
Furthermore, in the embodiment provided by the invention, the time relevance of the worksheets can be increased by combining the worksheets (the selected historical worksheets and the new worksheets) of each on-boarding attendant into the new worksheet sequence according to the time sequence, so that the characteristics of different on-boarding attendant are mined by using a trained deep learning model, and finally, the accuracy of pushing the worksheets can be effectively improved by reflecting the probability of the new worksheets of each on-boarding attendant.
Further, when the order pushing personnel are screened, the threshold is designated to screen the service personnel on the gate in the area where the new work order belongs, so that the situation that the improper service personnel on the gate push the work order can be effectively avoided, and the generation of interference information is reduced.
It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as a method, system, or computer program product. Accordingly, embodiments of the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, embodiments of the invention may take the form of a computer program product on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) having computer-usable program code embodied therein.
Embodiments of the present invention are described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
It will be apparent to those skilled in the art that various modifications and variations can be made to the present invention without departing from the spirit or scope of the invention. Thus, it is intended that the present invention also include such modifications and alterations insofar as they come within the scope of the appended claims or the equivalents thereof.

Claims (10)

1. A method of pushing a work order, comprising:
screening out regional gate service personnel in the region from all gate service personnel according to the region to which the new work order belongs;
the new worksheets are respectively built with the historical worksheets of each gate-on attendant in the regional gate-on attendant according to the time sequence, and a new worksheet sequence of each gate-on attendant is obtained; the historical worksheets are the specified number of worksheets completed by the corresponding worksheets receiving time of the new worksheets;
Based on each new work order sequence, calculating the new work order robbery probability of each gate-on service personnel for robbing the new work orders by using a trained deep learning model; the trained deep learning model is determined based on historical worksheets of all the service personnel;
screening the pushing staff based on all new work order robbing probabilities and specified thresholds corresponding to the area boarding staff; pushing the new work order to the order pushing personnel;
the method for obtaining the new work order sequence of each gate-on service person comprises the steps of:
preprocessing the new work order into data which can be identified by the trained deep learning model according to a specified rule, and obtaining a preprocessed new work order;
the new work order sequence of each door operator is constructed in the following way: starting to forward from the order receiving time corresponding to the new work order, and taking the preprocessed historical work orders corresponding to the appointed number of historical work orders of each gate-on service personnel from a database; the data stored in the database are the data of all the historical worksheets after being processed and the relation between the historical worksheets and service personnel; the preprocessed new worksheets and the appointed number of processed historical worksheets are assembled into a sequence according to the time sequence of completion, and a new worksheet sequence of each gate-up attendant is obtained; wherein the preprocessed new work order is the last element in the new work order sequence.
2. The method of claim 1, wherein before screening out regional gate service personnel in the region from all gate service personnel according to the region to which the new work order belongs, further comprising:
receiving order information of a user;
extracting basic information from the order information, and adding corresponding service information into the basic information to obtain the new work order; the basic information is contact information and address information of the user.
3. The method of claim 1, wherein preprocessing the new work order into data recognizable by the trained deep learning model according to a specified rule to obtain a preprocessed new work order comprises:
extracting feature data from the new work order to obtain a feature data set of the new work order; the type of the characteristic data is the type corresponding to the characteristic data used by each historical work order when the trained deep learning model is trained;
carrying out normalization processing on each characteristic data in the characteristic data set of the new work order to obtain a characteristic data set normalized by the new work order; the parameters adopted when the feature data are normalized are parameters adopted by the trained deep learning model corresponding to the types of the feature data during training;
And taking the characteristic data set normalized by the new work order as the preprocessed new work order.
4. The method of claim 1, wherein prior to calculating the new work order preemption probability for each gate service person to preempt the new work order using the trained deep learning model, further comprising:
training the deep learning model in the following manner to obtain the trained deep learning model:
preprocessing each historical work order into data which can be recognized by a deep learning model according to the appointed rule, and obtaining each preprocessed historical work order;
selecting a preset number of preprocessed historical worksheets for each gate-up service personnel, and constructing a historical worksheet sequence according to the time sequence in which the worksheets can be completed;
and inputting the historical work order sequences of all the service personnel to the deep learning model for repeated iterative learning until the training error of the deep learning model is smaller than a set threshold value, and obtaining the trained deep learning model.
5. The method of claim 1, wherein screening the order pushing personnel based on all new work order robbery probabilities and specified thresholds, comprises:
Calculating the specified threshold value by using a trained threshold value screening model based on a historical work order;
screening the new work order robbing probability of which the numerical value is not smaller than the specified threshold value from the new work order robbing probability to obtain screened new work order robbing probability;
and taking the service personnel on the gate corresponding to the screened new work order robbing probability as the order pushing personnel.
6. The method of claim 5, wherein calculating the specified threshold with a trained threshold screening model comprises:
extracting relevant information of each corresponding pushing person, each robbing person and acquiring person from the corresponding circulation information of each historical work order;
calculating the probability of each ticket pushing person, each ticket robbing person and the corresponding ticket robbing probability of the person corresponding to each historical work ticket through the trained deep learning model;
drawing all pushing worksheets according to all the worksheets pushing personnel, the worksheets grabbing personnel and the worksheets grabbing probabilities corresponding to the worksheets obtaining personnel, and obtaining worksheets grabbing probability distribution histograms corresponding to the worksheets;
selecting the probability corresponding to the highest histogram in the histograms corresponding to each type of characteristic data from the robbery probability histograms of all pushed worksheets and the robbery probability histograms of all obtained worksheets respectively to obtain a first threshold value of the pushed worksheets and a second threshold value of the obtained worksheets;
And selecting a minimum value from the first threshold value and the second threshold value as the specified threshold value.
7. The method of claim 5, wherein calculating the specified threshold with a trained threshold screening model comprises:
counting the number of pushing personnel pushing each historical work order;
clustering is carried out according to each characteristic data type of the work order, the number of pushing personnel of the historical work order under each characteristic data type is determined, and the number of pushing personnel corresponding to the characteristic data type with the largest number of pushing personnel in all the characteristic data types is selected as the specified threshold.
8. A system for pushing work orders, comprising:
the screening module is used for screening out regional boarding service personnel in the region from all boarding service personnel according to the region to which the new work order belongs;
the construction module is configured to construct the new worksheet with a historical worksheet of each on-boarding attendant in the regional on-boarding attendant in time sequence, and obtain a new worksheet sequence of each on-boarding attendant, where the construction module includes: preprocessing the new work order into data which can be recognized by a trained deep learning model according to a specified rule, and obtaining a preprocessed new work order; the new work order sequence of each door operator is constructed in the following way: starting to forward from the order receiving time corresponding to the new work order, and taking the preprocessed historical work orders corresponding to the appointed number of historical work orders of each gate-up service person from a database; the data stored in the database are the data of all the historical worksheets after being processed and the relation between the historical worksheets and service personnel; the pretreated new worksheets and the appointed number of treated historical worksheets are assembled into a sequence according to the time sequence of completion, and a new worksheet sequence of each door-to-door service person is obtained; the pretreated new work orders are the last element in the new work order sequence, the history work orders are the appointed number of work orders completed before the corresponding order receiving time of the new work orders for each gate-on service personnel;
The calculation module is used for calculating the new work order robbing probability of each gate-on service personnel for robbing the new work order by using the trained deep learning model based on each new work order sequence; the trained deep learning model is determined based on historical worksheets of all the service personnel;
the pushing module is used for screening the pushing personnel based on the probability of all new work orders corresponding to the regional gate-on service personnel and the specified threshold; and pushing the new work order to the order pushing personnel.
9. A system for pushing work orders, comprising:
at least one processor, and
a memory coupled to the at least one processor;
wherein the memory stores instructions executable by the at least one processor, the at least one processor performing the method of any of claims 1-7 by executing the instructions stored by the memory.
10. A computer-readable storage medium, characterized by:
the computer readable storage medium stores computer instructions which, when run on a computer, cause the computer to perform the method of any of claims 1-7.
CN201811329290.9A 2018-11-09 2018-11-09 Method, system and computer storage medium for pushing work order Active CN111178647B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201811329290.9A CN111178647B (en) 2018-11-09 2018-11-09 Method, system and computer storage medium for pushing work order

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201811329290.9A CN111178647B (en) 2018-11-09 2018-11-09 Method, system and computer storage medium for pushing work order

Publications (2)

Publication Number Publication Date
CN111178647A CN111178647A (en) 2020-05-19
CN111178647B true CN111178647B (en) 2023-09-05

Family

ID=70657028

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201811329290.9A Active CN111178647B (en) 2018-11-09 2018-11-09 Method, system and computer storage medium for pushing work order

Country Status (1)

Country Link
CN (1) CN111178647B (en)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115511397B (en) * 2022-11-23 2023-04-07 广东华居科技有限公司 Intelligent work order data dispatching method and system

Citations (13)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104954180A (en) * 2015-06-04 2015-09-30 广东省广播电视网络股份有限公司中山分公司 Broadcast television network comprehensive operation and maintenance supporting system and operation and maintenance method
CN105678524A (en) * 2016-03-03 2016-06-15 深圳棋致科技有限公司 Comprehensive service management method and apparatus for 4S store
CN106056318A (en) * 2016-07-12 2016-10-26 上海长城宽带网络服务有限公司 Service staff work distribution method based on broadband access and system
CN106296086A (en) * 2016-08-04 2017-01-04 温泉 A kind of businessman participates in the method and system of dispensing person's competition for orders
CN106845818A (en) * 2017-01-13 2017-06-13 北京市天元网络技术股份有限公司 Dispatching management information system based on broadband dress dimension
CN107154954A (en) * 2016-03-04 2017-09-12 苏州俊峰伟业信息技术有限公司 The row's of making house calls list method and system
CN107451878A (en) * 2017-06-01 2017-12-08 北京小度信息科技有限公司 order processing method and device
CN107808227A (en) * 2017-09-21 2018-03-16 大唐软件技术股份有限公司 A kind of distributing method and system of work order of constructing
WO2018058610A1 (en) * 2016-09-30 2018-04-05 深圳市华傲数据技术有限公司 Data collecting method and apparatus based on crowdsourcing, and server
WO2018072556A1 (en) * 2016-10-18 2018-04-26 无锡知谷网络科技有限公司 Logistics control method for goods, and electronic device
CN108090639A (en) * 2016-11-22 2018-05-29 上海安锐盟企业服务有限公司 Quality control system and method based on device service
CN108416490A (en) * 2018-01-15 2018-08-17 全球能源互联网研究院有限公司 A kind of electric communication operation and maintenance information-pushing method and system
CN108665170A (en) * 2018-05-14 2018-10-16 北京顺丰同城科技有限公司 Order allocation method and device

Family Cites Families (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20140101024A1 (en) * 2012-10-04 2014-04-10 Sap Ag Predicting financial outcome
WO2016019857A1 (en) * 2014-08-04 2016-02-11 北京嘀嘀无限科技发展有限公司 Service distribution system and method
US10402740B2 (en) * 2016-07-29 2019-09-03 Sap Se Natural interactive user interface using artificial intelligence and freeform input

Patent Citations (13)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104954180A (en) * 2015-06-04 2015-09-30 广东省广播电视网络股份有限公司中山分公司 Broadcast television network comprehensive operation and maintenance supporting system and operation and maintenance method
CN105678524A (en) * 2016-03-03 2016-06-15 深圳棋致科技有限公司 Comprehensive service management method and apparatus for 4S store
CN107154954A (en) * 2016-03-04 2017-09-12 苏州俊峰伟业信息技术有限公司 The row's of making house calls list method and system
CN106056318A (en) * 2016-07-12 2016-10-26 上海长城宽带网络服务有限公司 Service staff work distribution method based on broadband access and system
CN106296086A (en) * 2016-08-04 2017-01-04 温泉 A kind of businessman participates in the method and system of dispensing person's competition for orders
WO2018058610A1 (en) * 2016-09-30 2018-04-05 深圳市华傲数据技术有限公司 Data collecting method and apparatus based on crowdsourcing, and server
WO2018072556A1 (en) * 2016-10-18 2018-04-26 无锡知谷网络科技有限公司 Logistics control method for goods, and electronic device
CN108090639A (en) * 2016-11-22 2018-05-29 上海安锐盟企业服务有限公司 Quality control system and method based on device service
CN106845818A (en) * 2017-01-13 2017-06-13 北京市天元网络技术股份有限公司 Dispatching management information system based on broadband dress dimension
CN107451878A (en) * 2017-06-01 2017-12-08 北京小度信息科技有限公司 order processing method and device
CN107808227A (en) * 2017-09-21 2018-03-16 大唐软件技术股份有限公司 A kind of distributing method and system of work order of constructing
CN108416490A (en) * 2018-01-15 2018-08-17 全球能源互联网研究院有限公司 A kind of electric communication operation and maintenance information-pushing method and system
CN108665170A (en) * 2018-05-14 2018-10-16 北京顺丰同城科技有限公司 Order allocation method and device

Also Published As

Publication number Publication date
CN111178647A (en) 2020-05-19

Similar Documents

Publication Publication Date Title
CN110148285B (en) Intelligent oil well parameter early warning system based on big data technology and early warning method thereof
CN106600052B (en) User attribute and social network detection system based on space-time trajectory
CN107222865A (en) The communication swindle real-time detection method and system recognized based on suspicious actions
Lawrence et al. Improving a neural semantic parser by counterfactual learning from human bandit feedback
US20230385034A1 (en) Automated decision making using staged machine learning
CN104331502B (en) The recognition methods of courier's data in being marketed for courier periphery crowd
US11836582B2 (en) System and method of machine learning based deviation prediction and interconnected-metrics derivation for action recommendations
CN111260102A (en) User satisfaction prediction method and device, electronic equipment and storage medium
CN110110095A (en) A kind of power command text matching technique based on shot and long term memory Recognition with Recurrent Neural Network
CN105468161A (en) Instruction execution method and device
US20180196892A1 (en) Massively accelerated bayesian machine
CN107402859A (en) Software function verification system and verification method thereof
US10409914B2 (en) Continuous learning based semantic matching for textual samples
CN116010688A (en) User behavior label identification method, system, equipment and storage medium
CN111178647B (en) Method, system and computer storage medium for pushing work order
CN117094535B (en) Artificial intelligence-based energy supply management method and system
KR102223800B1 (en) System and methods thereof for providing smart bus line using artificial intelligence based on big data
CN108363738B (en) Recommendation method for industrial equipment data analysis algorithm
CN113807462A (en) AI-based network equipment fault reason positioning method and system
CN114117248A (en) Data processing method and device and electronic equipment
CN114627330A (en) Time sequence flow prediction method and device, storage medium and electronic equipment
CN116431319B (en) Task processing method and device
CN112669084B (en) Policy determination method, device and computer readable storage medium
US20230289746A1 (en) Methods and internet of things systems for creating smart gas call center work orders
WO2015132893A1 (en) Simulation analysis method and information processing system

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant