Customer intention intelligent identification and grading method

Abstract

The invention discloses a customer intention intelligent identification and grading method, and particularly relates to the technical field of intention grading. According to the method, three feature vectors of behaviors, interaction and attributes are constructed, multi-source data such as browsing tracks, consultation behaviors and demand preferences of clients are comprehensively covered, the one-sidedness of single-dimension evaluation is avoided, weighted calculation and standardized processing are combined, scattered data are converted into quantized intention evaluation coefficients, manual subjective judgment errors are greatly reduced, and the evaluation efficiency is improved. The client intention degree evaluation is enabled to better meet real requirements, and reliable data support is provided for accurate grading.

Description

Technical Field

[0001] This invention relates to the field of intent leveling technology, and more specifically, to a method for intelligent identification and leveling of customer intent. Background Technology

[0002] In fields such as real estate sales, online marketing, and customer relationship management, accurately identifying and classifying customer intentions is of great significance for improving sales conversion rates, optimizing resource allocation, and implementing precise follow-up.

[0003] Traditional customer intent assessment relies heavily on human experience, with sales staff subjectively judging the strength of customer intent and classifying and prioritizing customers accordingly. This method is not only inefficient but also highly subjective, inconsistent, and lagging, making it difficult to meet the real-time analysis needs of massive amounts of customer data.

[0004] With the development of information technology, some customer intention assessment methods based on rules or simple statistical models have emerged, but they still have the following shortcomings: The focus is too much on online customer behavior (such as browsing and clicking), and it fails to systematically integrate multi-source heterogeneous data such as offline customer interactions (such as proactive inquiries and appointments) and static attributes (such as customer profiles and demand matching), resulting in incomplete assessment profiles that are difficult to fully reflect the true intentions of customers. Furthermore, most assessments are based on data from a single point in time or a fixed period, failing to consider the dynamic trends of customer intentions over time. A customer with decreased recent interaction but consistently rising historical intentions may have drastically different potential value than a customer with frequent interaction but declining intentions. Static assessment methods cannot effectively capture such dynamic differences.

[0005] To address this, a smart customer intent recognition and grading method has been introduced. Summary of the Invention

[0006] To overcome the above-mentioned deficiencies of the prior art, embodiments of the present invention provide a method for intelligent identification and classification of customer intention.

[0007] To achieve the above objectives, the present invention provides the following technical solution: The intelligent identification and grading method for customer intent includes: S1: Construct a customer intent feature vector that includes behavioral features, interaction features, and attribute features; S2: After comprehensively processing the customer intention feature vector using pre-edited fusion processing logic, output the customer's intention evaluation coefficient in the current time zone; S3: Based on the mapping rule established between the intention assessment coefficient and the intention score, the customer's intention assessment coefficient in the current time zone is converted into an intention score, and the converted intention score is used as the customer's preliminary intention score in the current time zone. S4: Construct a customer intention sequence within the current evaluation period, analyze the changing trend of intention based on the customer intention sequence, and dynamically revise the initial intention score by combining the changing trend and the stage of customer follow-up to obtain the revised intention score. S5: Based on the corrected scores and feature vectors of all customers, unsupervised clustering analysis is performed to automatically divide the intention level clusters and integrate qualitative basis and core feature profiles for each level cluster.

[0008] Specifically, the logic for constructing behavioral features in S1; The number of times a customer visits a project page within the current time zone is counted as the browsing frequency. Identify the types of project pages visited by customers, count and sum the visit duration of project pages with the types of floor plan and price calculator, and use this as the core visit duration. Calculate the proportion of the core visit duration in the total visit duration of customers, and use this as the research depth. Locate the timestamp of the customer's most recent visit to the project page since the current time, and calculate the time difference between the current time and the most recent timestamp of the visit to the project page as the recent activity level. Identify the number of queries a customer makes while accessing a project page within the current time zone, and categorize the number of queries by query type into school district queries, transportation queries, and related queries. The number of times a customer queries school districts, transportation, and related services within the current time zone is multiplied by the set weights for school districts, transportation, and related services, respectively, and then summed to determine the content preference level. The behavioral characteristics are output after weighted calculation logic is used to comprehensively process the customer's browsing frequency, research depth, activity level, and content preference in the current time zone.

[0009] Specifically, the logic for constructing interactive behavior features in S1; The number of inquiries initiated by customers within the current time zone is counted as the number of inquiries initiated. The percentage of the number of inquiries initiated to the total number of interactions within the current time zone is calculated as the inquiry ratio. Extract the customer's interaction content during the interaction process in the current time zone, input the interaction content into a pre-built project keyword library for matching, and count the number of successfully matched project keyword types as the professionalism of the question; The number of times a customer schedules a viewing within the current time zone is counted. The ratio of the number of scheduling appointments to the number of actual attendances is used to calculate the viewing reference score. The interaction characteristics are output after weighted calculation logic is used to comprehensively process the customer's proactive consultation rate, question professionalism, and viewing participation in the current time zone.

[0010] Specifically, the attribute feature construction process in S1; Define the feature vector M of the project as [ , , , ..., ], where M includes the feature number of the item in each dimension, and x is the total number of dimensions; Collect customer dimensional preferences and construct a preference vector C=[ , , , ..., ]; where C includes the expected dimensions proposed by the customer, and j is the total number of expected dimensions proposed; Match the project feature vector M and the customer preference vector C, count the number of dimensions in the project feature vector M that satisfy the customer preference vector as the satisfaction quantity, and calculate the proportion of the satisfaction quantity in the total number of expected dimensions as the attribute feature.

[0011] Specifically, S2 outputs the customer's intention assessment coefficient in the current time zone; Extract behavioral features, interaction features, and attribute features from the customer intent feature vector; After standardization, a weighted calculation logic is used to obtain the customer's intention evaluation coefficient in the current time zone.

[0012] Specifically, S4 analyzes the changing trends of intent based on customer intent sequences; The customer intention sequence is constructed by arranging the intention evaluation coefficients of X historical time zone customers and the intention evaluation coefficients of current time zone customers in chronological order. Extract the intention evaluation coefficients of two adjacent groups to construct an evaluation pair. Subtract the intention evaluation coefficient on the left from the intention evaluation coefficient on the right in the evaluation pair to obtain the evaluation difference. If the assessment difference is positive, then mark the current two adjacent groups of intention assessment coefficients as intention increase values; The total number of intention increases in the customer intention sequence is counted as the number of increases. The percentage of rising numbers in the total number of assessments is used to obtain the intended trend ratio.

[0013] Specifically, S4 incorporates the current stage of customer follow-up; Starting from the time of the customer's first interaction, the time difference between the starting point and the current time is calculated to obtain the customer's follow-up duration. The follow-up duration is matched with three preset follow-up duration intervals. Each follow-up duration interval corresponds to a follow-up stage, and each follow-up stage corresponds to a follow-up adjustment coefficient. The follow-up stages include the contact stage, the communication stage, and the pre-sale stage.

[0014] Specifically, in S4, the initial intention score is dynamically revised; Calculate the difference between the intention trend ratio and 50%, and mark the difference as the score amplification value if the difference is positive. If the difference is negative, then the difference is marked as the score reduction value; Based on the labeling results, the pre-established mapping rules for rating amplification value-amplification factor or rating reduction value-reduction factor are selectively used to convert the amplification factor or reduction factor. Using formula Receive customer's revision intention rating ;in A preliminary score is given to the customer's intention in the current time zone. gt represents the magnification factor or the reduction factor. t=1 or 2. t=1 is the magnification factor, and t=2 is the reduction factor. To follow up and adjust the coefficients.

[0015] Specifically, S5 categorizes customers into various interest level clusters; Define the intention rating range corresponding to different level clusters, match the customer's revised intention rating with the corresponding intention rating range, and output the customer's intention level cluster.

[0016] Specifically, S5 integrates qualitative criteria and core feature profiles for each level cluster; The customer's intention evaluation coefficient is used as the qualitative basis for the grade cluster; Behavioral features, interaction features, and attribute features are extracted from the feature vector. After standardization, three line segments are extended from the origin at equal angles. The extension length of the three line segments corresponds to the specific values ​​of the standardized behavioral features, interaction features, and attribute features, respectively. After the extension is completed, the extension endpoints of the three line segments are connected in sequence to construct a triangle. The constructed triangle is used as the core feature profile.

[0017] The technical effects and advantages of this invention are as follows: (1) By constructing three feature vectors of behavior, interaction and attributes, we can fully cover multi-source data such as customer browsing trajectory, consultation behavior and demand preferences, avoid the one-sidedness of single-dimensional evaluation, and combine weighted calculation and standardized processing to transform scattered data into quantitative intention evaluation coefficients, greatly reduce the error of human subjective judgment, make customer intention evaluation more in line with real needs, and provide reliable data support for accurate classification. (2) By constructing a customer intention sequence based on historical and current data, capturing the intention fluctuation pattern through intention trend ratio analysis, and combining the adjustment coefficients corresponding to the follow-up stage, the initial score is dynamically corrected. This avoids the shortcomings of static evaluation in failing to reflect changes in customer intentions, and can adapt the scoring weights according to different stages such as contact, communication, and pending transaction, so that the grading results are in line with the customer status in real time, improving the timeliness and adaptability of the grading. (3) Automatically divide the four intention level clusters through unsupervised clustering, and present the core characteristics of each level of customers in a triangular feature profile to help the business team quickly locate high-intent customers and prioritize the allocation of resources for precise follow-up; formulate personalized cultivation or maintenance strategies for different levels of customers to avoid the waste of marketing resources and significantly improve customer conversion efficiency and operational resource utilization efficiency. Attached Figure Description

[0018] Figure 1 This is a flowchart of the intelligent identification and grading method for customer intent of the present invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] like Figure 1 As shown, the intelligent identification and grading method for customer intent includes: Intent feature construction: Collect multi-source heterogeneous data related to customers, including behavioral data, interaction data, and attribute data; Behavioral data: Customer interaction sequences and timestamps on the digital display platform of the property, including: number of times and duration of visits to the project details page, frequency and duration of viewing floor plans and VR model rooms, number of times and calculation parameters of the price calculator, repeated access records to key information pages such as school district, transportation, and supporting facilities, as well as collection and sharing behaviors; Interaction data: Multi-channel communication content and metadata between customers and the enterprise, including: dialogue records from sales offices or online consultations, text transcriptions of WeChat or telephone communications, email consultation content, communication details for scheduling property viewings; as well as communication time, communication duration, and response interval; Attribute data: Customer background profile information, including: family structure, work location and commuting needs, current housing situation, budget range, home purchase eligibility, and loan qualification pre-approval status.

[0021] The collected data is preprocessed, including data cleaning, missing value handling, outlier handling, word segmentation of unstructured text data, and removal of stop words.

[0022] Based on the preprocessed data, construct the customer's intention feature vector in the current time zone, including behavioral features, interaction features, and attribute features; The behavioral characteristics are derived from a comprehensive analysis of the customer's browsing frequency, research depth, activity level, and content preference. Behavioral feature construction logic; The number of times a customer visits a project page within the current time zone is counted as the browsing frequency. During the process of counting page views, visits with a page dwell time of less than 3 seconds were excluded (accidental clicks, crawler crawling). Exclude duplicate requests (refresh actions) from the same IP address and device ID accessing the same item page consecutively within 1 minute.

[0023] Identify the types of project pages visited by customers, count and sum the visit duration of project pages with the types of floor plan and price calculator, and use this as the core visit duration. Calculate the proportion of the core visit duration in the total visit duration of customers, and use this as the research depth. Research depth definition: Core access time ÷ Total access time (sum of time spent on all valid pages) × 100%.

[0024] Extract the floor plan page and price calculator page records from the valid records: For example, the core access duration within the current time zone is extracted as follows: (1) Floor plan page: 45 seconds; (2) House price calculator (60 seconds); (3) Floor plan page: 30 seconds; (4) Floor plan page: 25 seconds; (5) House price calculator (90 seconds); (5) Floor plan page: 50 seconds; 45 + 60 + 30 + 25 + 90 + 50 = 300 seconds (5 minutes); Assuming the total visit duration (the sum of the dwell time of all valid records) is 435, the research depth is calculated as follows: (300÷435)×100%≈68.97%→Research depth≈69%.

[0025] Locate the timestamp of the customer's most recent visit to the project page since the current time, and calculate the time difference between the current time and the most recent timestamp of the visit to the project page as the recent activity level. Identify the number of queries a customer makes while accessing a project page within the current time zone, and categorize the number of queries by query type into school district queries, transportation queries, and related queries. The number of times a customer queries school districts, transportation, and related services within the current time zone is multiplied by the set weights for school districts, transportation, and related services, respectively, and then summed to determine the content preference level. Assume the number of times school district queries, transportation queries, and related queries are 2, 1, and 2 respectively; Content preference level = (Number of school district queries × School district weight) + (Number of transportation queries × Transportation weight) + (Number of related queries × Related weight); Preset weights (for first-time homebuyer properties in the real estate industry, adjustable according to project positioning): School district weight = 0.4, Transportation weight = 0.3, Amenities weight = 0.3; Content preference score = (2×0.4) + (1×0.3) + (2×0.3) = 0.8 + 0.3 + 0.6 = 1.7 points; the maximum number of queries for a single type is 5.

[0026] The behavioral characteristics are output after comprehensive processing of the customer's browsing frequency, research depth, activity level, and content preference in the current time zone using weighted calculation logic. The behavioral feature output process is as follows: The customer's browsing frequency, research depth, activity level, and content preference within the current time zone are respectively marked as follows: ; Based on a pre-defined set of behavioral references, the formula is used: Behavioral characteristics are obtained after calculation ;in , , , All of these are set weighting coefficients; This is a behavioral reference set, representing the frequency of reference browsing, the depth of reference research, the level of reference activity, and the degree of preference for reference content, respectively. The behavioral reference set is set based on the average customer behavior of similar projects in the real estate industry (such as affordable housing, upgrade housing, and luxury housing) to ensure that the threshold conforms to the general rules of the industry. Alternatively, it can be based on the project's own historical data to explore the behavioral characteristic thresholds of "converted customers" (such as the average browsing frequency of converted customers ≥ 6 times and the research depth ≥ 50%). The passing score is set with "conversion relevance" as the core. During the setting process, the threshold needs to be dynamically adjusted according to the significant differences in customer behavior characteristics at different stages (such as the passing score can be lowered when the frequency of customer behavior is low in the introductory stage and increased when customer behavior is stable in the mature stage).

[0027] Interaction characteristics are derived from a comprehensive analysis of the customer's proactive inquiry rate, the professionalism of their questions, and their participation in the viewing process. Logic for constructing interactive behavior features; The number of inquiries initiated by customers within the current time zone is counted as the number of inquiries initiated. The percentage of the number of inquiries initiated to the total number of interactions within the current time zone is calculated as the inquiry ratio. The number of inquiries initiated by customers (such as online customer service inquiries, telephone inquiries, message inquiries, etc., excluding content passively pushed by the system / sales); Total number of interactions: All interactions between the client and the project team within the current time zone (proactive inquiries + reactive responses).

[0028] Extract the customer's interaction content during the interaction process in the current time zone, input the interaction content into a pre-built project keyword library for matching, and count the number of successfully matched project keyword types as the professionalism of the question; Project Keyword Library: Presets keyword types and terms related to the core attributes of the project (examples below); Calculation logic: Match the interactive content of customer inquiries with the keyword database and count the number of successfully matched "keyword types" (multiple matches of the same type are counted only once). Example of a project keyword library: (1) Usable floor area ratio: usable floor area ratio, actual usable area, and shared area; (2) Building parameters: floor height, building spacing, plot ratio, greening rate; (3) Building material standards: building material brands, decoration standards, door and window materials, wall coatings; (4) Enrollment policies: school district division, enrollment requirements, designated schools, and household registration policies; (5) Property-related fees, property management company, and scope of property services; For example, a customer's interaction might be: "What is the approximate usable floor area ratio of this project?" → This contains the keyword "usable floor area ratio" → Matching type: "usable floor area ratio".

[0029] The number of times a customer schedules a viewing within the current time zone is counted. The ratio of the number of scheduling appointments to the number of actual attendances is used to calculate the viewing reference score. Number of appointments: The total number of times customers have proactively made appointments to view the project within the current time zone (including online appointments, telephone appointments, etc.); Actual attendance: The number of times the property was actually viewed at the project site after the appointment was made; Special note: If the actual number of visits is 0, the reference value of the viewing is 0; if the number of appointments is 0, the reference value of the viewing is 0.

[0030] The interaction characteristics are output after comprehensive processing of the customer’s proactive consultation rate, question professionalism and viewing participation in the current time zone using weighted calculation logic. The interactive feature output process is as follows: The customer's proactive inquiry rate, the professionalism of their questions, and their engagement in viewings within the current time zone are respectively categorized as follows: ; Combining a pre-defined interactive reference set, using the formula: Behavioral characteristics are obtained after calculation ;in , , All of these are set weighting coefficients; This is an interactive reference set, representing the ratio of proactive inquiries to references, the professionalism of reference questions, and the participation rate in referral viewings, respectively.

[0031] The attribute feature is the degree of matching between the project feature vector and the customer preference vector; Attribute feature construction process; Define the feature vector M of the project as [ , , , ..., ], where M includes the feature number of the item in each dimension, and x is the total number of dimensions; Additional dimension descriptions, such as: - Apartment type matching degree; - School district quality levels, including general, district key, and city key; - Greening the environment; ......; -Delivery standards include fully furnished, partially furnished, and unfurnished.

[0032] Collect customer dimensional preferences and construct a preference vector C=[ , , , ..., ]; where C includes the expected dimensions proposed by the customer, and j is the total number of expected dimensions proposed; Match the project feature vector M with the customer preference vector C, count the number of dimensions in the project feature vector M that satisfy the customer preference vector as the satisfaction quantity, and calculate the proportion of the satisfaction quantity in the total number of expected dimensions as the attribute feature. Following the principle of "one-to-one matching between customer expectations and corresponding project dimensions," determine whether project characteristics meet customer expectations, and count the number of dimensions that are met. Example illustration: - The school district's quality level is at least a key district level. If the focus is on a specific area, then customer preferences will be met.

[0033] Preliminary Intention Score: After comprehensively processing the customer's intention feature vector using pre-edited fusion processing logic, the customer's intention evaluation coefficient in the current time zone is output. Based on the mapping rule established between the intention evaluation coefficient and the intention score, the customer's intention evaluation coefficient in the current time zone is converted into an intention score, and the converted intention score is used as the customer's preliminary intention score in the current time zone. The comprehensive processing is as follows: Extracting behavioral features from customer intent feature vectors Interaction features and attribute characteristics ; All samples were standardized; the standardization process used Min-Max normalization; the formula was used... The customer's intention evaluation coefficient in the current time zone is calculated. ;in , as well as The weighting coefficients are set. Based on the mapping rule established between the intention assessment coefficient and the intention score, the customer's intention assessment coefficient in the current time zone is converted into an intention score, and the converted intention score is used as the customer's preliminary intention score in the current time zone. That is, by setting the range of each group of coefficients corresponding to the intention assessment coefficient, and each group of coefficients corresponds to an intention score; the intention score range is limited to 1-100, and the larger the intention assessment coefficient, the higher the probability of matching 100. Explanation of the interval and corresponding intention rating example: Interval 1: 0 < ≤0.2 corresponds to an intention score of 1-20; Interval 2: 0.2 < ≤0.45 corresponds to an intention score of 21-45; Interval 3: 0.45 < ≤0.65 corresponds to an intention score of 46-65; Interval 4: 0.65 < ≤0.85 corresponds to an intention score of 66-85; Interval 5: 0.85 < ≤1.0 corresponds to an intention score of 86-100; Assuming the intention evaluation coefficient is 0.929, corresponding to the range of 0.85-1.0, the score is 86-100, the range length is 0.15, and the score span is 14; Then the intention score = 86 + ( -0.85) / 0.15×14; Calculate: 86 + (0.929 - 0.85) / 0.15 × 14 ≈ 86 + 7.37 ≈ 93.37 points (rounded to 93 points).

[0034] Real-time dynamic correction: Extract the intention evaluation coefficients of customers from the previous X historical time zones in the current time zone, construct the customer intention sequence within the current evaluation period, analyze the changing trend of intention based on the customer intention sequence, and dynamically correct the initial intention score by combining the changing trend and the stage of customer follow-up to obtain the corrected intention score; X>3, the specific value is set by technical personnel, and can be dynamically corrected according to the actual situation in the future. Specifically: The customer intention sequence is constructed by arranging the intention evaluation coefficients of X historical time zone customers and the intention evaluation coefficients of current time zone customers in chronological order. Extract the intention evaluation coefficients of two adjacent groups to construct an evaluation pair. Subtract the intention evaluation coefficient on the left from the intention evaluation coefficient on the right in the evaluation pair to obtain the evaluation difference. If the assessment difference is positive, then mark the current two adjacent groups of intention assessment coefficients as intention increase values; The total number of intention increases in the customer intention sequence is counted as the number of increases. Calculate the percentage of rising numbers in the total number of assessment pairs to obtain the intention trend ratio; Calculate the difference between the intention trend ratio and 50%, and mark the difference as the score amplification value if the difference is positive. If the difference is negative, then the difference is marked as the score reduction value; Based on the labeling results, the pre-established mapping rules for rating amplification value-amplification factor or rating reduction value-reduction factor are selectively used to convert the amplification factor or reduction factor. Additional explanation of mapping rules: That is, by setting a score amplification value, there are amplification value ranges corresponding to each set of amplification values, and each set of amplification value ranges corresponds to a set of amplification coefficients; the amplification coefficient range is limited to 1.05-1.18, and the larger the score amplification value, the higher the probability of matching 1.18; The score reduction value is preset to correspond to each set of reduction value intervals, and each set of reduction value intervals corresponds to a set of reduction coefficients; the reduction coefficient range is limited to 0.891-0.954, and the smaller the score reduction value, the higher the probability of matching 0.891; Explained using the distance between magnified value intervals: Mark the amplification value as ; Interval 1: 0 < If the magnification is ≤10%, then the corresponding amplification factor is 1.05; Interval 2: 10% < If the magnification is ≤30%, then the corresponding amplification factor is 1.09; Interval 1: 30% < If the magnification is ≤40%, then the corresponding magnification factor is 1.13; Interval 1: 40% < If the magnification is ≤100%, the corresponding magnification factor is 1.18.

[0035] Starting from the time of the customer's first interaction, the time difference between the starting point and the current time is calculated to obtain the customer's follow-up duration. The follow-up duration is matched with three preset follow-up duration intervals. Each follow-up duration interval corresponds to a follow-up stage, and each follow-up stage corresponds to a follow-up adjustment coefficient. The follow-up stages include the contact stage, the communication stage, and the pre-sale stage. The follow-up adjustment coefficient is limited to the range of 0.961-1.086, and the follow-up adjustment coefficient for the pre-sale stage is greater than that for the communication stage, which is greater than that for the contact stage. These coefficients can be set to 1.086, 1.014, and 0.961, respectively. The follow-up time range can be preset, for example: Contact phase: 0-21 days; Communication phase: 21-60 days; Closing phase: 60 days or more.

[0036] Using formula Receive customer's revision intention rating ;in A preliminary score is given to the customer's intention in the current time zone. gt represents the magnification factor or the reduction factor. t=1 or 2. t=1 is the magnification factor, and t=2 is the reduction factor. To adjust the coefficients accordingly, the result is rounded to the nearest integer.

[0037] Intention grading results: Unsupervised clustering analysis was performed on the revised intention scores and feature vectors of all customers to automatically divide customers into various intention level clusters; qualitative criteria and core feature profiles were integrated for each level cluster, which included high intention urgent type, medium intention under observation type, low intention nurturing type, and no intention type. Specifically: Define the intention rating range corresponding to different level clusters, match the customer's revised intention rating with the corresponding intention rating range, and output the customer's intention level cluster. Example of intention rating range: Urgent Intent Type: 86-100 points. If a customer's revised intent score is above 100, then the default score is 100. Medium-term assessment type: 46-85 points; Low intention to cultivate: 11-45 points; Uninterested type: 1-10 points.

[0038] The customer's intention evaluation coefficient is used as the qualitative basis for the grade cluster; Behavioral features, interaction features, and attribute features are extracted from the feature vector. After standardization, three line segments are extended from the origin at equal angles. The extension length of the three line segments corresponds to the specific values ​​of the standardized behavioral features, interaction features, and attribute features, respectively. After the extension is completed, the extension endpoints of the three line segments are connected in sequence to construct a triangle. The constructed triangle is used as the core feature profile. The above formulas are all dimensionless calculations. Dimensionless calculations can be performed using various methods such as standardization, which will not be elaborated here. The formulas are derived from software simulations based on a large amount of collected data, and the preset parameters in the formulas can be set by those skilled in the art according to the actual situation.

[0039] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, ATA hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. The semiconductor medium can be a solid-state ATA hard disk.

[0040] It should be understood that in the various embodiments of this application, the order of the above-mentioned processes does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0041] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0042] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0043] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0044] In addition, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0045] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable ATA hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0046] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for intelligent identification and grading of customer intention, characterized in that, include: S1: Construct a customer intent feature vector that includes behavioral features, interaction features, and attribute features; S2: After comprehensively processing the customer intention feature vector using pre-edited fusion processing logic, output the customer's intention evaluation coefficient in the current time zone; S3: Based on the mapping rule established between the intention assessment coefficient and the intention score, the customer's intention assessment coefficient in the current time zone is converted into an intention score, and the converted intention score is used as the customer's preliminary intention score in the current time zone. S4: Construct a customer intention sequence within the current evaluation period, analyze the changing trend of intention based on the customer intention sequence, and dynamically revise the initial intention score by combining the changing trend and the stage of customer follow-up to obtain the revised intention score. S5: Based on the corrected scores and feature vectors of all customers, unsupervised clustering analysis is performed to automatically divide the intention level clusters and integrate qualitative basis and core feature profiles for each level cluster.

2. The intelligent identification and grading method for customer intention as described in claim 1, characterized in that: The logic for constructing behavioral features in S1; The number of times a customer visits a project page within the current time zone is counted as the browsing frequency. Identify the types of project pages visited by customers, count and sum the visit duration of project pages with the types of floor plan and price calculator, and use this as the core visit duration. Calculate the proportion of the core visit duration in the total visit duration of customers, and use this as the research depth. Locate the timestamp of the customer's most recent visit to the project page since the current time, and calculate the time difference between the current time and the most recent timestamp of the visit to the project page as the recent activity level. Identify the number of queries a customer makes while accessing a project page within the current time zone, and categorize the number of queries by query type into school district queries, transportation queries, and related queries. The number of times a customer queries school districts, transportation, and related services within the current time zone is multiplied by the set weights for school districts, transportation, and related services, respectively, and then summed to determine the content preference level. The behavioral characteristics are output after weighted calculation logic is used to comprehensively process the customer's browsing frequency, research depth, activity level, and content preference in the current time zone.

3. The intelligent identification and grading method for customer intention as described in claim 1, characterized in that: The logic for constructing interactive behavior features in S1; The number of inquiries initiated by customers within the current time zone is counted as the number of inquiries initiated. The percentage of the number of inquiries initiated to the total number of interactions within the current time zone is calculated as the inquiry ratio. Extract the customer's interaction content during the interaction process in the current time zone, input the interaction content into a pre-built project keyword library for matching, and count the number of successfully matched project keyword types as the professionalism of the question; The number of times a customer schedules a viewing within the current time zone is counted. The ratio of the number of scheduling appointments to the number of actual attendances is used to calculate the viewing reference score. The interaction characteristics are output after weighted calculation logic is used to comprehensively process the customer's proactive consultation rate, question professionalism, and viewing participation in the current time zone.

4. The intelligent identification and grading method for customer intention as described in claim 1, characterized in that: The attribute feature construction process in S1; Define the feature vector M of the project as [ , , , ..., ], where M includes the feature number of the item in each dimension, and x is the total number of dimensions; Collect customer dimensional preferences and construct a preference vector C=[ , , , ..., ]; where C includes the expected dimensions proposed by the customer, and j is the total number of expected dimensions proposed; Match the project feature vector M and the customer preference vector C, count the number of dimensions in the project feature vector M that satisfy the customer preference vector as the satisfaction quantity, and calculate the proportion of the satisfaction quantity in the total number of expected dimensions as the attribute feature.

5. The intelligent identification and grading method for customer intention as described in claim 1, characterized in that: S2 outputs the customer's intention assessment coefficient in the current time zone; Extract behavioral features, interaction features, and attribute features from the customer intent feature vector; After standardization, a weighted calculation logic is used to obtain the customer's intention evaluation coefficient in the current time zone.

6. The intelligent identification and grading method for customer intention as described in claim 1, characterized in that: S4 analyzes the changing trend of intention based on customer intention sequence; The customer intention sequence is constructed by arranging the intention evaluation coefficients of X historical time zone customers and the intention evaluation coefficients of current time zone customers in chronological order. Extract the intention evaluation coefficients of two adjacent groups to construct an evaluation pair. Subtract the intention evaluation coefficient on the left from the intention evaluation coefficient on the right in the evaluation pair to obtain the evaluation difference. If the assessment difference is positive, then mark the current two adjacent groups of intention assessment coefficients as intention increase values; The total number of intention increases in the customer intention sequence is counted as the number of increases. The percentage of rising numbers in the total number of assessments is used to obtain the intended trend ratio.

7. The intelligent identification and grading method for customer intention as described in claim 6, characterized in that: S4 incorporates the current stage of customer follow-up; Starting from the time of the customer's first interaction, the time difference between the starting point and the current time is calculated to obtain the customer's follow-up duration. The follow-up duration is matched with three preset follow-up duration intervals. Each follow-up duration interval corresponds to a follow-up stage, and each follow-up stage corresponds to a follow-up adjustment coefficient. The follow-up stages include the contact stage, the communication stage, and the pre-sale stage.

8. The intelligent identification and grading method for customer intention as described in claim 7, characterized in that: S4 dynamically corrects the initial intention score; Calculate the difference between the intention trend ratio and 50%, and mark the difference as the score amplification value if the difference is positive. If the difference is negative, then the difference is marked as the score reduction value; Based on the labeling results, the pre-established mapping rules for rating amplification value-amplification factor or rating reduction value-reduction factor are selectively used to convert the amplification factor or reduction factor. Using formula Receive customer's revision intention rating ;in A preliminary score is given to the customer's intention in the current time zone. gt represents the magnification factor or the reduction factor. t=1 or 2. t=1 is the magnification factor, and t=2 is the reduction factor. To follow up and adjust the coefficients.

9. The intelligent identification and grading method for customer intention as described in claim 8, characterized in that: In S5, customers are divided into various intention level clusters; Define the intention rating range corresponding to different level clusters, match the customer's revised intention rating with the corresponding intention rating range, and output the customer's intention level cluster.

10. The intelligent identification and grading method for customer intention as described in claim 1, characterized in that... S5 integrates qualitative criteria and core feature profiles for each level cluster; The customer's intention evaluation coefficient is used as the qualitative basis for the grade cluster; Behavioral features, interaction features, and attribute features are extracted from the feature vector. After standardization, three line segments are extended from the origin at equal angles. The extension length of the three line segments corresponds to the specific values ​​of the standardized behavioral features, interaction features, and attribute features, respectively. After the extension is completed, the extension endpoints of the three line segments are connected in sequence to construct a triangle. The constructed triangle is used as the core feature profile.

Images