CN114866651B - Node-reusable intelligent outbound method and system - Google Patents
Node-reusable intelligent outbound method and system Download PDFInfo
- Publication number
- CN114866651B CN114866651B CN202210791358.5A CN202210791358A CN114866651B CN 114866651 B CN114866651 B CN 114866651B CN 202210791358 A CN202210791358 A CN 202210791358A CN 114866651 B CN114866651 B CN 114866651B
- Authority
- CN
- China
- Prior art keywords
- user
- node
- time
- curve
- users
- 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
Links
- 238000000034 method Methods 0.000 title claims abstract description 34
- 239000013598 vector Substances 0.000 claims abstract description 36
- 238000006243 chemical reaction Methods 0.000 claims abstract description 17
- 238000012163 sequencing technique Methods 0.000 claims abstract description 13
- 230000006399 behavior Effects 0.000 claims description 47
- 238000004364 calculation method Methods 0.000 claims description 13
- 230000004044 response Effects 0.000 claims description 7
- 238000004590 computer program Methods 0.000 claims description 6
- 230000006870 function Effects 0.000 claims description 6
- 238000013139 quantization Methods 0.000 claims description 4
- 230000008859 change Effects 0.000 claims description 2
- 239000003795 chemical substances by application Substances 0.000 abstract description 6
- 238000012545 processing Methods 0.000 abstract description 6
- 238000012216 screening Methods 0.000 abstract 1
- 230000007246 mechanism Effects 0.000 description 4
- 238000004458 analytical method Methods 0.000 description 3
- 238000004891 communication Methods 0.000 description 3
- 230000035484 reaction time Effects 0.000 description 3
- 230000009471 action Effects 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 239000002699 waste material Substances 0.000 description 2
- 238000012952 Resampling Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 238000004422 calculation algorithm Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000014509 gene expression Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 208000024891 symptom Diseases 0.000 description 1
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04M—TELEPHONIC COMMUNICATION
- H04M3/00—Automatic or semi-automatic exchanges
- H04M3/42—Systems providing special services or facilities to subscribers
- H04M3/50—Centralised arrangements for answering calls; Centralised arrangements for recording messages for absent or busy subscribers ; Centralised arrangements for recording messages
- H04M3/51—Centralised call answering arrangements requiring operator intervention, e.g. call or contact centers for telemarketing
- H04M3/5175—Call or contact centers supervision arrangements
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04M—TELEPHONIC COMMUNICATION
- H04M3/00—Automatic or semi-automatic exchanges
- H04M3/42—Systems providing special services or facilities to subscribers
- H04M3/50—Centralised arrangements for answering calls; Centralised arrangements for recording messages for absent or busy subscribers ; Centralised arrangements for recording messages
- H04M3/51—Centralised call answering arrangements requiring operator intervention, e.g. call or contact centers for telemarketing
- H04M3/523—Centralised call answering arrangements requiring operator intervention, e.g. call or contact centers for telemarketing with call distribution or queueing
- H04M3/5232—Call distribution algorithms
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04M—TELEPHONIC COMMUNICATION
- H04M3/00—Automatic or semi-automatic exchanges
- H04M3/42—Systems providing special services or facilities to subscribers
- H04M3/50—Centralised arrangements for answering calls; Centralised arrangements for recording messages for absent or busy subscribers ; Centralised arrangements for recording messages
- H04M3/51—Centralised call answering arrangements requiring operator intervention, e.g. call or contact centers for telemarketing
- H04M3/523—Centralised call answering arrangements requiring operator intervention, e.g. call or contact centers for telemarketing with call distribution or queueing
- H04M3/5238—Centralised call answering arrangements requiring operator intervention, e.g. call or contact centers for telemarketing with call distribution or queueing with waiting time or load prediction arrangements
-
- 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
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Landscapes
- Engineering & Computer Science (AREA)
- Signal Processing (AREA)
- Marketing (AREA)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention relates to the technical field of data processing, in particular to a node reusable intelligent outbound method and a node reusable intelligent outbound system. The method comprises the steps of obtaining word frequency vectors and problem node propulsion curves of user filled information, and constructing a user appeal behavior package; classifying users based on the user appeal behavior package to obtain a plurality of user groups; acquiring the identity of the consumed time of problem nodes between two groups of user groups, and screening out multiplexing problem nodes; acquiring a head propulsion curve of each user based on the multiplexing node; based on any user, calculating the affinity corresponding to the head propulsion curve of the user in the head propulsion curve user group to obtain a queuing factor of each user; and carrying out outbound sequencing on the users based on the queuing factor. The invention divides the users to obtain the queuing factors of the users by analyzing the word frequency vectors and the key reaction conditions of the users, performs outbound sequencing on the users, and realizes the purposes of intelligently allocating manual agents, saving agent resources and avoiding wasting user time.
Description
Technical Field
The invention relates to the technical field of data processing, in particular to a node reusable intelligent outbound method and a node reusable intelligent outbound system.
Background
A call-out reservation type call-out has become a flexible technique for improving the quality of service and the experience of a client, such as technical support, call-back complaint, etc., the call-out reservation requires the client to participate to a certain extent in advance, the call-out work order is formed by a form of information filling such as a questionnaire, etc., and then an artificial seat is assigned to communicate with the client. The appointment calling type can meet the client requirements more accurately and humanizedly based on the service required by the client self-definition and the delivery time.
Because the reserved call-out type is large in group size, multiple in types and relatively subdivided in personalized service content, how to effectively distribute the robot nodes avoids wasting user time, saves limited seat resources and improves communication efficiency of users is a problem that optimization is needed at present.
Disclosure of Invention
In order to solve the above technical problems, an object of the present invention is to provide a method and a system for node-reusable intelligent outbound, wherein the adopted technical scheme is as follows:
in a first aspect, an embodiment of the present invention provides an intelligent outbound method that a node can multiplex, where the method includes the following steps:
acquiring a word frequency vector of user filled information; analyzing the key behavior of each problem node heard by a user after the user switches on the voice, and recording the problem level and the response time to obtain a problem node push curve; constructing a user appeal behavior package based on the word frequency vector and the problem node propulsion curve;
classifying a plurality of users based on the similarity degree of the user appeal behavior package to obtain a plurality of user groups;
for any group of user groups, obtaining the time consumption consistency of each problem node based on the consumption time of each problem node in the user groups; the similarity of time consumption consistency of problem nodes of corresponding problem levels of any two groups of user groups is obtained and used as identity, and the problem node corresponding to the minimum identity in each user group is selected and used as a multiplexing problem node;
acquiring problem node propelling curves corresponding to all problem nodes before each user reaches the multiplexing problem node as head propelling curves; based on any user, calculating affinity corresponding to a head propulsion curve of the user in the user group and the head propulsion curve of the user in the user group, wherein the affinity is used as a queuing factor; and carrying out outbound sequencing on the users based on the queuing factor.
Preferably, the obtaining the word frequency vector of the user filling information includes:
and calculating the user filling information by using a CountVectorizer function to obtain a word frequency vector.
Preferably, the classifying a plurality of users based on the similarity degree of the user appeal behavior package includes:
calculating the spatial distance of the user appeal behavior packet corresponding to each user, and classifying the plurality of users based on the spatial distance;
the calculation formula of the spatial distance is as follows:
wherein,
the spatial distance of the user appeal behavior package between the user p and the other user q;
advancing a curve for a problem node corresponding to the pth user;
advancing a curve for a problem node corresponding to the qth user;
the word frequency vector corresponding to the p-th user;
a word frequency vector corresponding to the qth user;
the reaction duration of the p-th user; the reaction duration of the qth user;
advancing curves to problem nodes
And problem node push curve
The covariance of (a); advancing curves to problem nodes
And problem node push curve
Standard deviation of (2).
Preferably, the obtaining of the time consumption consistency of each question node based on the time consumption of each question node in the user group includes:
acquiring a corresponding time-consuming time set of each user based on each problem node; and calculating the standard deviation of the time-consuming time set of each problem node, adding one to the standard deviation to serve as a time-consuming index, and taking the reciprocal of the time-consuming index as the time-consuming consistency of the problem nodes.
Preferably, the obtaining of the similarity of the time-consuming consistency of the problem nodes of the corresponding problem hierarchy of any two groups of user groups as the identity includes:
sequencing the time consumption consistency from large to small, and selecting Top-K time consumption consistency; calculating similarity of the time consumption consistency among different user groups as identity based on Top-K time consumption consistency;
the calculation formula of the identity is as follows:
wherein,
the identity of the kth problem node in Top-K different problem nodes in different user groups is obtained; n is the number of other user groups except the user group to which the current user belongs; x is the index number of the problem node; time-consuming consistency of the xth problem node in the user group to which the current user belongs;
time-consuming consistency of the xth problem node in the user group except the current user; abs () is an absolute value function.
Preferably, the calculating the affinity of the head propulsion curve corresponding to the head propulsion curve of the user in the user group includes:
calculating the initial affinity of any two users in the same user group;
the calculation formula of the initial affinity is as follows:
wherein,
initial affinity corresponding to the user p and the other user q;
a head push curve corresponding to the pth user;
a head push curve corresponding to the qth user;
and calculating the reciprocal of the initial affinity mean value between the user and other users in the corresponding user group, and taking the reciprocal as the affinity of the user.
In a second aspect, an embodiment of the present invention provides a node-reusable intelligent outbound system, including a memory, a processor, and a computer program stored in the memory and executable on the processor, where the processor implements the node-reusable intelligent outbound method when executing the computer program.
The embodiment of the invention at least has the following beneficial effects:
the embodiment of the invention utilizes a data processing technology, and the method utilizes the data processing technology and obtains the word frequency vector of the user filling information; analyzing the key behavior of each problem node heard by a user after the user switches on the voice, and recording the problem level and the response time to obtain a problem node propulsion curve; constructing a user appeal behavior package based on the word frequency vector and the problem node propulsion curve; classifying a plurality of users based on the similarity degree of the user appeal behavior package to obtain a plurality of user groups; for any group of user groups, obtaining the time consumption consistency of each problem node based on the consumption time of each problem node in the user groups; the similarity of time consumption consistency of problem nodes of corresponding problem levels of any two groups of user groups is obtained as identity, and a problem corresponding to the minimum identity in each user group is selected as a multiplexing problem node; acquiring problem node propelling curves corresponding to all problem nodes before each user reaches the multiplexing problem node as head propelling curves; based on any user, calculating the affinity of the head propelling curve corresponding to the head propelling curves of the users in the user group, and taking the affinity as a queuing factor; and carrying out outbound sequencing on the users based on the queuing factor. According to the invention, the user is analyzed through the word frequency vector and the key reaction condition of the user to obtain the queuing factor of the user, and the user is subjected to outbound sequencing based on the queuing factor, so that the purposes of intelligently allocating artificial agents, saving agent resources and avoiding wasting user time are realized.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions and advantages of the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present invention, and other drawings can be obtained by those skilled in the art without creative efforts.
Fig. 1 is a flowchart of a method for a node-reusable intelligent outbound method according to an embodiment of the present invention.
Detailed Description
To further illustrate the technical means and effects of the present invention adopted to achieve the predetermined objects, the following detailed description will be given to a node-reusable intelligent outbound method and system according to the present invention, and the detailed implementation, structure, features and effects thereof with reference to the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" refers to not necessarily the same embodiment. Furthermore, the particular features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.
Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.
The embodiment of the invention provides a node reusable intelligent outbound method and a specific implementation method of a system, and the method is suitable for an intelligent outbound queuing scene. In the scene, the outbound call reservation user needs to fill in questionnaires first to reserve the call time of the outbound call system, and the user can perform key operation according to voice prompt after the voice is switched on. In order to solve the problems of seat resource waste and low user communication efficiency, the invention analyzes the user to obtain the user queuing factor by the word frequency vector and the key reaction condition of the user, and performs outbound sequencing on the user based on the queuing factor, thereby realizing the purposes of intelligently allocating manual seats, saving seat resources and avoiding user time waste.
The following describes a specific scheme of the node-reusable intelligent outbound method and system provided by the invention in detail with reference to the accompanying drawings.
Referring to fig. 1, a flowchart illustrating steps of a node-reusable intelligent outbound method according to an embodiment of the present invention is shown, where the method includes the following steps:
step S100, acquiring a word frequency vector of user filled information; analyzing the key behavior of each problem node heard by a user after the user switches on the voice, and recording the problem level and the response time to obtain a problem node push curve; and constructing a user appeal behavior package based on the word frequency vector and the problem node propulsion curve.
The service content of the outbound call is various, the user selects a large class of problems based on text forms such as questionnaire filling, the calling time of the outbound call system is reserved, and the characteristics of the filling information and the user appeal characteristics of one user can be reflected through the behaviors of the robot problem nodes after the outbound call.
And analyzing semantic features of the user filled information, acquiring word frequency vectors filled by the user, analyzing the key behavior of each problem heard by the user after the user puts through voice after calling out, namely analyzing the behavior of the user through the node based on the key behavior of voice prompt after calling out, and constructing a user appeal behavior package. It should be noted that each question, i.e., each voice prompt, serves as a question node.
Specifically, the method comprises the following steps: for all records from the receipt of an outbound call to the eventual arrival of manual service by each of the robotic question nodes, the following analysis is performed:
firstly, semantic features of filled information are analyzed, and word frequency vectors filled by users are obtained.
For general filling information, the main content is concise and brief, and generally comprises problem symptoms, problem keywords and the like.
Firstly, based on all filling information from system operation to date, text word frequency statistics based on a word bag model is carried out on the filling information, after the word frequency statistics is carried out by the word bag model, the word frequency of words of the filling information which can be seen by the outbound service can be obtained, and an implementer only needs to eliminate common words and nonsense words which repeatedly appear in the field due to expressions. At present, there are many word banks commonly used in the field, and the implementer can select the excluded words more flexibly.
For a filled-in information text, after word segmentation, the bag-of-word model can obtain word-based features of the text, namely a word vector, by counting the frequency of each word in the text.
Further, a countvectorer function is used for calculating the filling information to obtain a word frequency vector F.
Step two, analyzing the key behavior of each question heard by the user after the user switches on the voice, recording the question hierarchy and the response time to obtain a question node advancing curve
And analyzing the key-press behaviors of the user at each problem node, wherein the key-press selection behavior is performed from the root node of the binary or multi-branch tree to the end node of the final problem logic step by step. It should be noted that, for the user key operation without going to manual, recording and analysis are not performed. Therefore, no matter what kind of problem nodes, the problem nodes of each level are deepened step by step and finally reach the manual customer service. Therefore, the time overlapped by voice broadcast, reaction, thinking and action based on the question asked by the user to the system is counted, and finally the key behavior characteristics of the user are obtained:
first, taking a problem node as an example, after hearing the voice broadcast, the user may have the following reactions:
(1) the questions are known, the answers are known, and the numbers are selected at the moment when the voice broadcast starts.
(2) The questions are unknown and the answers are known, selection is carried out when the voice broadcast reaches the keywords or the question, the voice broadcast is interrupted in advance, and the next question is entered.
(3) And in other cases that the questions are unknown and the answers are unknown, the selection is carried out after the time longer than the voice broadcasting time.
Because the problem depth is deepened along with the time, the user can reflect the behavior characteristics of the user in the key event reaction condition of a section of problem node. Based on the above, the problem level-reaction time change of the user is recorded, and a problem node advancing curve is obtained.
The generation mode of the problem node propulsion curve is as follows: and taking 1s as quantization time, and recording the problem node where the user is currently located.
As a critical user who may have been able to predict how much time will be wasted on the robot next, the situation within seconds of the phone call being on may be as follows: where each number represents the problem level at which the customer was at the current sampling time. Wherein 0 is manual, and the duration of each question level in the sequence is the reaction time of the current question node user. It can be known that the user has a short response time and almost selects the user based on the pre-judgment problem and the key number.
Since the information amount of each problem node is different, the reaction time of the user is also different, but the following indexes can be obtained:
reaction duration of interaction time belonging to robot node
And relative problem node propulsion curves in the time period
。
And performing nearest neighbor resampling on the problem node propulsion curve, and scaling the curve time to preset quantization duration, so as to normalize the characteristics represented by the reaction duration of each problem node. And updating the resampled problem node advancing curve to serve as a problem node advancing curve. In the embodiment of the present invention, the preset quantization duration is 40, and in other embodiments, an implementer may adjust the value according to an actual situation.
The problem node based push curve can represent a time consumption characteristic that a user pushes the problem node to the end when the user is at the robot node, and therefore the user with different key behaviors can be separated based on the difference of push curve characteristics and duration.
Step three, constructing a user appeal behavior package based on the word frequency vector and the problem node propulsion curve
Wherein, F is a word frequency vector, and S is a problem node propulsion curve. The user appeal behavior package is constructed based on the semantic features and the key behavior features.
Different appeal, understanding degree of the appeal and urgency degree of the user can be distinguished based on the obtained user appeal behavior packet.
And S200, classifying the plurality of users based on the similarity of the user appeal behavior packages to obtain a plurality of user groups.
Unsupervised classification is carried out on the appeal of different users, user groups with similar user appeal are automatically determined,
the improved DBSCAN clustering algorithm based on the user appeal behavior package classifies user groups of users by combining time domain characteristics and text information characteristics of behaviors, an abstract group type can be constructed for all the user groups under the unsupervised condition, the similarity degree based on the user appeal behavior package is realized, a plurality of users are classified, and a plurality of user groups are obtained.
Whether the problem node propulsion curves in different user requirements are similar or not is restrained based on semantic features of the filling information, the user behaviors can not be analyzed aiming at the problem nodes, the problem semantic ambiguity is avoided, and meanwhile the problem library content of the user implementing the outbound call system is avoided being accessed. Design of
As the spatial distance in clustering.
The calculation formula of the spatial distance of the user appeal behavior package between the arbitrary user p and the another user q is as follows:
wherein,
advancing a curve for a problem node corresponding to the pth user;
advancing a curve for a problem node corresponding to the qth user;
the word frequency vector corresponding to the p-th user;
a word frequency vector corresponding to the qth user; the reaction duration of the p-th user;
the reaction duration of the qth user;
advancing curves to problem nodes
And problem node push curve
The covariance of (a);
problem node push curve
And problem node push curve
Is the standard deviation.
Wherein, in the space distance formula
Is a coefficient of determination of behavior characteristics with a value range of [0,1 ]]If the similarity degree between the user behavior and the semantics is not high, the coefficient tends to 1 so as to enlarge the difference between the reaction times and further distinguish the users.
Problem node push curve after time length normalization
If the propulsion speeds between nodes are similar and the durations are similar, they behave similarly regardless of appeal.
For the problems with different durations and similar propulsion speeds, the corresponding problems are likely to be different, because the meaning of the user to the problem node can be determined only after the key problem vocabulary is heard, and if the whole problem has a large deviation, such as 10s, the problem faced by the user is different.
For time length similarity but high determinant coefficient, the relevance of the problem node advancing curve is weak, and the word vector similarity is weak. Directly amplified by the time length difference:
if the time lengths are similar, the condition is sporadic, and a lower queuing factor is still given in the subsequent analysis.
For time lengths that are quite dissimilar, they are treated directly as outliers. Even if a type is formed, the later intimacy is weak, and a higher queuing factor cannot be obtained.
And classifying the plurality of users based on the appeal and the key behavior to obtain user group categories of the plurality of user groups.
Step S300, for any group of user groups, obtaining the time consumption consistency of each problem node based on the consumption time of each problem node in the user groups; and obtaining the similarity of the time consumption consistency of the problem nodes of the corresponding problem levels of any two groups of user groups as identity, and selecting the problem node corresponding to the minimum identity in each user group as a multiplexing problem node.
And calculating the effectiveness and semantic applicability of the problem node to the user group, and taking the problem node with lower applicability under the user group as a multiplexing problem node. Namely, each problem is taken as a node, and the multiplexing node is selected according to the effectiveness and the applicability of the problem node to the user group. Specifically, the method comprises the following steps:
since some problems may cause a certain user group to pay low attention to the problem, or the problem is too useless or even redundant for the group to choose, in order to find such a problem node, the semantic applicability of the problem node is calculated:
firstly, a group of user groups is determined, and the uniformity of problem voice playing progress of each problem of the user groups is calculated:
and constructing a feature vector J for the time consumed by each user problem node, wherein the feature vector is a feature vector formed by the time consumed by the problem node.
Even if the correlation of the problem node propulsion curve can be extremely strong to restrict the problems in the group of the type to be consistent, the problem can occur in individual cases in one group, and the problems are not uniform among other users and different in number, so that the maximum problem depth in the user group is based on. Calculating a feature vector J in the population, and constructing a new m-dimensional hypothesis space so as to analyze the behavior consistency of each question: the time-consuming distribution of each problem node is analyzed.
For each problem node, a time-consuming set of time per user can be obtained
And i is the ith node, and further calculating the time-consuming consistency of each node.
Calculating the standard deviation of each time-consuming time set, and adding one to the standard deviation to serve as a time-consuming index; the inverse of the time consumption indicator is the time consumption consistency of the problem node.
Time-consuming consistency R of the ith problem node i The calculation formula of (2) is as follows:
wherein, the standard deviation of the time consuming set of the ith problem node is shown.
Sequencing the time consumption consistency from large to small to obtain a time consumption consistency sequence, and selecting Top-K time consumption consistency; and calculating the similarity of the time-consuming consistency among different user groups as the identity based on the Top-K time-consuming consistency. Namely, the problem with high time-consuming consistency of Top-K calculation is compared with the time-consuming consistency of the problem nodes of the rest user groups. It should be noted that the implementer can appropriately mask the problem, such as the first problem, according to the situation of the problem node. Among them, Top-K is used for analyzing the problem of large time-consuming consistency of extraction.
Each problem node gets the index number IDX of the problem, wherein the index number of the problem is the order of Top-K in the time-consuming consistency sequence.
Time-consuming consistency based on Top-K questions
And inquiring the corresponding time-consuming consistency of another user group:
。
calculating the identity of the kth problem node in Top-K different problem nodes in different user groups
：
Wherein N is exceptThe number of other user groups outside the user group to which the current user belongs; x is the index number of the question;
time-consuming consistency of the xth problem node in the user group to which the current user belongs;
time-consuming consistency of the xth problem node in the user group except the current user; abs () is an absolute value function. And selecting the problem node corresponding to the minimum group identity in each user group as a multiplexing problem node, namely a multiplexing node.
When the group identity is high, since each problem node is carefully designed, there is a high probability that the problem node is important in the entire outbound query node, otherwise the group is considered insignificant to the problem.
Step S400, acquiring problem node propelling curves corresponding to all problem nodes before each user reaches the multiplexing problem node as head propelling curves; based on any user, calculating the affinity of the head propelling curve corresponding to the head propelling curves of the users in the user group, wherein the affinity is used as a queuing factor; and carrying out outbound sequencing on the users based on the queuing factor.
Triggering a pre-queuing mechanism based on a multiplexing node, distributing the nearest user group in real time when a user performs dialing operation, and dynamically distributing the current queuing factor Q based on the condition of continuously dialing answer questions.
And based on the behavior record of the user group, temporarily taking the multiplexing node as a final problem level, and performing secondary calculation on the problem node propelling curve of the group to obtain the problem node propelling curves corresponding to all the problem nodes before each user reaches the multiplexing problem node as a head propelling curve.
Further, a queuing factor of the user is obtained according to the similarity of the head propelling curve.
When the user encounters the multiplexing node, calculating the corresponding head propulsion curve and the belonged userAffinity of the population: given the joint distance calculation method, assuming that the user has entered the nth node, there is a head push curve before the user reaches the nth node
。
Initial affinity corresponding to arbitrary user p and another user q
The calculation formula of (2) is as follows:
wherein,
a head push curve corresponding to the pth user;
the curve is advanced for the corresponding head of the qth user.
And calculating the reciprocal of the initial affinity mean value between the user and other users in the corresponding user population, and taking the reciprocal as affinity. That is, there is an initial affinity that the user has K neighbors:
formula for calculating affinity
Comprises the following steps:
k is the number of users in a user group to which the current user belongs;
all user samples in a user group to which the user belongs; initial affinity for user p and another user q。
The system intercepts the problem node propulsion curve of the history record before reaching the nth node corresponding to all the obtained user groups, and can construct a cache in advance by the mechanism so as to accelerate the calculation.
This affinity is taken as a queuing factor. Based on the queuing factor, the users are sequenced for outbound call, that is, the users are queued, called, or switched to another user group.
When the queuing factor is smaller than the preset queuing threshold, the user continues to return to a subsequent problem node based on the multiplexing node, and is possibly switched to another user group.
And when the queuing factor is larger than or equal to the preset queuing threshold, ranking is rearranged based on the current online waiting user and the queuing factor, namely, a dynamic queuing mechanism is implemented, and if the queue is short, the user is directly assigned to manual customer service when the user continues to answer the question, so that the experience is improved.
And when the queuing factor is continuously smaller than the preset queuing threshold value, directly distributing manual customer service to the user.
Namely, when the user does not walk to the final artificial channel, the system dynamically allocates a queuing factor to the user.
Based on the mechanism, when the user receives the outgoing call, the system can compare the user with the group result of the current clustering of the system based on the key-press behavior of the user and the text characteristic before the outgoing call is initiated, and analyze based on the most similar group.
When the user presses the key, the problem node enters the next stage, and the previous problem node advancing curve is used as a head advancing curve:
updating the most similar group of the user, if the user reaches the multiplexing node, trying to queue based on the current queuing factor, and if no one answers, waiting for answering when the user continues to operate; if the user does not reach the multiplexing node, queuing is not needed.
The reason for the low user queuing factor is mainly influenced by: the user knows the options to a high degree, and is likely to need to communicate with customer service frequently, so that the waiting time can be greatly shortened by using a queuing factor; the user has less knowledge of the problem, meaning that it may not belong to any category, nor is it known to the problem node because of frequent communication, i.e. it may face an unanswered, possibly helpful, problem option, and therefore even though the group's multiplex nodes are passed, it still queues up later without entering the last manual option.
In summary, the embodiment of the present invention utilizes a data processing technique, and the method obtains the word frequency vector of the user filled information; analyzing the key behavior of each problem node heard by a user after the user switches on the voice, and recording the problem level and the response time to obtain a problem node push curve; constructing a user appeal behavior package based on the word frequency vector and the problem node propulsion curve; classifying a plurality of users based on the similarity degree of the user appeal behavior package to obtain a plurality of user groups; for any group of user groups, obtaining the time consumption consistency of each problem based on the consumption time of each problem node in the user groups; the similarity of time consumption consistency of problem nodes of corresponding problem levels of any two groups of user groups is obtained and used as identity, and the problem node corresponding to the minimum identity in each user group is selected and used as a multiplexing problem node; acquiring problem node propelling curves corresponding to all problem nodes before each user reaches the multiplexing problem node as head propelling curves; based on any user, calculating the affinity of the head propelling curve corresponding to the head propelling curves of the users in the user group, and taking the affinity as a queuing factor; and carrying out outbound sequencing on the users based on the queuing factor. According to the invention, the user is analyzed through the word frequency vector and the key reaction condition of the user to obtain the queuing factor of the user, and the user is subjected to outbound sequencing based on the queuing factor, so that the purposes of intelligently allocating artificial agents, saving agent resources and avoiding wasting user time are realized.
The embodiment of the invention also provides an intelligent outbound system with reusable nodes, which comprises a memory, a processor and a computer program which is stored in the memory and can run on the processor, wherein the processor realizes the steps of the method when executing the computer program. Since the detailed description of the reusable intelligent outbound method of a node is given above, it is not described again.
It should be noted that: the precedence order of the above embodiments of the present invention is only for description, and does not represent the merits of the embodiments. And specific embodiments thereof have been described above. Other embodiments are within the scope of the following claims. In some cases, the actions or steps recited in the claims may be performed in a different order than in the embodiments and still achieve desirable results. In addition, the processes depicted in the accompanying figures do not necessarily require the particular order shown, or sequential order, to achieve desirable results. In some embodiments, multitasking and parallel processing may also be possible or may be advantageous.
The embodiments in the present specification are described in a progressive manner, and the same and similar parts among the embodiments are referred to each other, and each embodiment focuses on the differences from the other embodiments. The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents, improvements and the like that fall within the spirit and principle of the present invention are intended to be included therein.
Claims (6)
1. An intelligent outbound method reusable by a node, characterized by comprising the following steps:
acquiring a word frequency vector of user filled information; analyzing the key behavior of each problem node heard by a user after the user switches on the voice, and recording the problem level and the response time to obtain a problem node push curve; constructing a user appeal behavior package based on the word frequency vector and the problem node propulsion curve;
classifying a plurality of users based on the similarity degree of the user appeal behavior package to obtain a plurality of user groups;
for any group of user groups, obtaining the time consumption consistency of each problem node based on the consumption time of each problem node in the user groups; the similarity of time consumption consistency of the problem nodes of the corresponding problem levels of any two groups of user groups is obtained and used as identity, and the problem node corresponding to the minimum identity in each user group is selected as a multiplexing problem node;
acquiring problem node propelling curves corresponding to all problem nodes before each user reaches the multiplexing problem node as head propelling curves; based on any user, calculating affinity of a head propulsion curve corresponding to the head propulsion curves of the users in the user group, wherein the affinity is used as a queuing factor; based on the queuing factor, carrying out outbound sequencing on the users;
the problem node advancing curve is obtained according to the change of the problem hierarchy-reaction time of the user, and 1s is used as quantization time to record the problem node where the user is located currently;
the method for calculating the affinity of the head propulsion curve corresponding to the head propulsion curve of the users in the user group comprises the following steps: calculating the initial affinity of any two users in the same user group;
the calculation formula of the initial affinity is as follows:
wherein,
initial affinity corresponding to the user p and the other user q;
a head push curve corresponding to the pth user;
a head push curve corresponding to the qth user;
and calculating the reciprocal of the initial affinity mean value between the user and other users in the corresponding user group, and taking the reciprocal as the affinity of the user.
2. The node-reusable intelligent outbound method according to claim 1, wherein said obtaining a word-frequency vector of the user-filled information comprises:
and calculating the user filling information by using a CountVectorizer function to obtain a word frequency vector.
3. The node-reusable intelligent outbound method of claim 1, wherein said classifying a plurality of users based on the similarity of said user appeal behavior package comprises:
calculating the spatial distance of the user appeal behavior packet corresponding to each user, and classifying the plurality of users based on the spatial distance;
the calculation formula of the spatial distance is as follows:
wherein,
the spatial distance of the user appeal behavior package between the user p and the other user q;
advancing a curve for a problem node corresponding to the pth user;
advancing a curve for a problem node corresponding to the qth user;
the word frequency vector corresponding to the p-th user;
a word frequency vector corresponding to the qth user;
the reaction duration of the p-th user;
the reaction duration of the qth user;
advancing curves to problem nodes
And problem node push curve
The covariance of (a);
advancing curves to problem nodes
And problem node push curve
Standard deviation of (2).
4. The node-reusable intelligent outbound method according to claim 1, wherein said obtaining time-consuming consistency of each problem node based on the time-consuming of each problem node in the user population comprises:
acquiring a corresponding time-consuming time set of each user based on each problem node; and calculating the standard deviation of the time-consuming time set of each problem node, adding one to the standard deviation to serve as a time-consuming index, and taking the reciprocal of the time-consuming index as the time-consuming consistency of the problem nodes.
5. The node-reusable intelligent outbound method according to claim 1, wherein said obtaining similarity of time-consuming consistency of problem nodes of corresponding problem levels of any two groups of user groups as identity comprises:
sequencing the time consumption consistency from large to small, and selecting Top-K time consumption consistency; calculating similarity of the time consumption consistency among different user groups as identity based on Top-K time consumption consistency;
the calculation formula of the identity is as follows:
wherein,
the identity of the kth problem node in Top-K different problem nodes in different user groups is obtained; n is the number of other user groups except the user group to which the current user belongs; x is the index number of the problem node;
time-consuming consistency of the xth problem node in the user group to which the current user belongs;
time-consuming consistency of the xth problem node in the user group except the current user; abs () is an absolute value function.
6. A node-reusable intelligent outbound system comprising a memory, a processor and a computer program stored in said memory and executable on said processor, characterized in that said processor implements the steps of the method according to any one of claims 1 to 5 when executing said computer program.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210791358.5A CN114866651B (en) | 2022-07-07 | 2022-07-07 | Node-reusable intelligent outbound method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210791358.5A CN114866651B (en) | 2022-07-07 | 2022-07-07 | Node-reusable intelligent outbound method and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114866651A CN114866651A (en) | 2022-08-05 |
| CN114866651B true CN114866651B (en) | 2022-09-20 |
Family
ID=82626617
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210791358.5A Active CN114866651B (en) | 2022-07-07 | 2022-07-07 | Node-reusable intelligent outbound method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114866651B (en) |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2000005867A1 (en) * | 1998-07-22 | 2000-02-03 | British Telecommunications Public Limited Company | Call queuing in a telecommunications network |
| CN111787171A (en) * | 2020-06-24 | 2020-10-16 | 中国平安财产保险股份有限公司 | Automatic outbound method, device, electronic equipment and computer readable storage medium |
| CN111967885A (en) * | 2020-08-18 | 2020-11-20 | 中国银行股份有限公司 | Intelligent outbound call processing method and device |
| CN112163686A (en) * | 2020-10-23 | 2021-01-01 | 江西师范大学 | Waiting response time confidence interval prediction method and device for G/G/1 queuing system |
| CN113163063A (en) * | 2021-04-30 | 2021-07-23 | 重庆度小满优扬科技有限公司 | Intelligent outbound call system and method |
| CN113873087A (en) * | 2021-09-29 | 2021-12-31 | 中国工商银行股份有限公司 | Outbound method, device, computer equipment and storage medium |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110392168B (en) * | 2018-04-16 | 2021-06-01 | 华为技术有限公司 | Call processing method, device, server, storage medium and system |
-
2022
- 2022-07-07 CN CN202210791358.5A patent/CN114866651B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2000005867A1 (en) * | 1998-07-22 | 2000-02-03 | British Telecommunications Public Limited Company | Call queuing in a telecommunications network |
| CN111787171A (en) * | 2020-06-24 | 2020-10-16 | 中国平安财产保险股份有限公司 | Automatic outbound method, device, electronic equipment and computer readable storage medium |
| CN111967885A (en) * | 2020-08-18 | 2020-11-20 | 中国银行股份有限公司 | Intelligent outbound call processing method and device |
| CN112163686A (en) * | 2020-10-23 | 2021-01-01 | 江西师范大学 | Waiting response time confidence interval prediction method and device for G/G/1 queuing system |
| CN113163063A (en) * | 2021-04-30 | 2021-07-23 | 重庆度小满优扬科技有限公司 | Intelligent outbound call system and method |
| CN113873087A (en) * | 2021-09-29 | 2021-12-31 | 中国工商银行股份有限公司 | Outbound method, device, computer equipment and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114866651A (en) | 2022-08-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US7013005B2 (en) | System and method for prioritizing contacts | |
| KR102381214B1 (en) | Man-machine dialogue method, device and electronic equipment | |
| US11165723B2 (en) | Methods and systems for a bimodal auto-response mechanism for messaging applications | |
| KR102121135B1 (en) | Optimized routing of interactions to contact center agents based on machine learning | |
| CN110263144A (en) | A kind of answer acquisition methods and device | |
| US7245706B2 (en) | Voice response system | |
| US7406413B2 (en) | Method and system for the processing of voice data and for the recognition of a language | |
| CN108153800A (en) | Information processing method, information processing unit and program | |
| CN110059182A (en) | Art recommended method and device towards customer service | |
| MX2011002272A (en) | Call routing methods and systems based on multiple variable standardized scoring and shadow queue. | |
| US20020188681A1 (en) | Method and system for informing users of subjects of discussion in on -line chats | |
| US20040002868A1 (en) | Method and system for the processing of voice data and the classification of calls | |
| CN115982331A (en) | Information interaction method, device and equipment in session scene | |
| CN114760387A (en) | Method and device for managing maintenance | |
| CN114866651B (en) | Node-reusable intelligent outbound method and system | |
| US20120023173A1 (en) | System and method for prioritizing message transcriptions | |
| CN114047995A (en) | Method, device and equipment for determining label color and storage medium | |
| CN112015879A (en) | Man-machine interaction engine implementation method and device based on text structured management | |
| US20040006464A1 (en) | Method and system for the processing of voice data by means of voice recognition and frequency analysis | |
| Jagerman et al. | Models and approximations for call center design | |
| US7343288B2 (en) | Method and system for the processing and storing of voice information and corresponding timeline information | |
| CN113743956A (en) | System and method for on-line customer service instant chat | |
| Hathaway | Data-driven Studies of Caller Behavior Under Call Center Innovations | |
| CN115348356A (en) | Queuing voice information processing method and device, computer equipment and storage medium | |
| CN113744741B (en) | Method and device for recognizing telephone answering behavior |
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 |