An Event Intelligent Detection Method and System for an AI Electronic Employee Badge

Abstract

The present invention relates to the technical field of event detection, and specifically discloses an event intelligent detection method and system for an AI electronic work card, including: collecting, identifying, and classifying events based on the AI electronic work card; performing frequency analysis on customer event categories to determine frequent categories; quantitatively analyzing the cycle time of frequent categories to determine event-intensive sub-cycle times, and analyzing the periodic correlation between the company-triggered event time and the event-intensive sub-cycle times through an autocorrelation function; if there is a correlation, quantitatively analyzing the processing ability of the wearer of the AI electronic work card to determine the preferred person with the same ability; constructing an event category prediction model to predict the event-intensive sub-cycle times and customer event categories, and arranging the preferred person with the same ability to handle them. The present invention can intelligently detect events, analyze event patterns, evaluate personnel capabilities, and perform reasonable task allocation, improving the efficiency and quality of event handling.

Description

Technical Field

[0001] The present invention relates to the technical field of event detection, and particularly to an event intelligent detection method and system for an AI electronic work card. Background Art

[0002] Event detection technology plays a key role in improving the level of enterprise customer interaction management. However, the existing event detection technology is difficult to meet the increasing refined management needs of enterprises.

[0003] A Chinese patent application with the publication number CN114462418B discloses an event detection method, system, intelligent terminal and computer-readable storage medium, including: obtaining a statement to be detected and an event category sequence; generating data to be detected by combining the statement to be detected with the event category sequence; obtaining the self-attention model and the multi-layer perceptron, and obtaining the event category probability corresponding to the statement to be detected based on the self-attention model and the multi-layer perceptron.

[0004] In the prior art, the frequency analysis of customer event categories is not carried out, and the frequent categories are not determined. If the customer event categories are obtained, analyzed by using chi-square test and Poisson distribution, and then combined with the deviation data set to calculate the event frequency value, it is beneficial to distinguish the frequently occurring and occasional event categories, so that enterprises have a basis for resource allocation and reduce the possibility of resource waste or insufficient allocation.

[0005] In the prior art, the quantitative analysis of the cycle time of frequent categories is not carried out, and the association between the company-triggered event time and the event-intensive sub-cycle time is not established. If the curve distribution is judged and then the periodic association is determined by the autocorrelation function analysis method, enterprises can make preparations in advance based on the periodic association and improve the business control ability.

[0006] In the prior art, the quantitative analysis of the ability of AI work card wearers to handle events is not carried out, and the personnel are not reasonably arranged according to the event prediction results. If the method of determining the periodic association, then quantitatively analyzing the processing ability of AI electronic work card wearers, constructing an event category prediction model, and arranging the best-performing personnel of the same type to handle the corresponding events is adopted, the problems of low success rate and efficiency of enterprise event handling can be solved, which is beneficial to improving the event handling efficiency and quality, and enabling enterprises to plan resources in advance and formulate coping strategies.

[0007] Therefore, the present invention provides an event intelligent detection method and system for an AI electronic work card. Summary of the Invention

[0008] The purpose of the present invention is to provide an event intelligent detection method and system for an AI electronic work card to solve the problems in the above background.

[0009] The object of the present invention can be achieved by the following technical solutions:

[0010] An event intelligent detection method for an AI electronic work card, comprising the following steps:

[0011] Collect audio and scene pictures, identify events and classify the events, and construct an event classification table;

[0012] Based on the event classification table, perform a frequency analysis on the categories of customer events deviating from the data set to obtain frequent categories;

[0013] Perform a quantitative analysis on the cycle time of the frequent categories to obtain the time of the event-intensive sub-cycle, obtain the time when the company triggers an event, and determine whether there is a periodic correlation between the time when the company triggers an event and the time of the event-intensive sub-cycle through statistical analysis;

[0014] If there is a periodic correlation, perform a quantitative analysis on the processing ability of the AI electronic work card wearer for customer events of frequent categories, determine the preferred ability persons of the same category, and construct an ability person data set;

[0015] Construct an event category prediction model, predict the time of the event-intensive sub-cycle and the category of customer events, and arrange the preferred ability persons of the same category to the predicted event category of the prediction sub-cycle.

[0016] As a further solution of the present invention: the determination method of the frequent category is:

[0017] During the monitoring period, based on the event classification table, obtain customer events of different categories and construct a customer event data set;

[0018] Based on the customer event data set, use the chi-square test to determine whether the customer event data of different categories in the customer event data set conforms to the Poisson distribution;

[0019] If it conforms, split the customer event data set into a fitting data set, otherwise, split it into a deviating data set;

[0020] Through the Poisson distribution probability calculation formula, perform a probability analysis on the fitting data set to determine the frequent category;

[0021] Perform numerical calculation on the deviating data set to determine the frequent category.

[0022] As a further solution of the present invention: the method for performing numerical calculation on the deviating data set to determine the frequent category is:

[0023] Based on the deviating data set, obtain the number of consecutive occurrences and the number of interval occurrences of the same customer event category during the monitoring period;

[0024] Obtain the event frequency value through a weighted formula based on the consecutive occurrence times and interval times within the monitoring period;

[0025] If the event frequency value of customer events of different categories is higher than the preset frequency threshold, mark the customer event category as a frequent category.

[0026] As a further solution of the present invention: The determination method of whether there is a periodic association is as follows:

[0027] Based on the customer events of the frequent category, obtain the number of customer events within multiple sub-periods of the monitoring period, and draw a monitoring period - event number change curve;

[0028] Perform statistical analysis on the monitoring period - event number change curve to obtain a time difference sequence;

[0029] Based on the time difference sequence, determine whether there is a periodic association between the time when the company triggers an event and the time of the event-intensive sub-period through the autocorrelation function analysis method and the chi-square distribution test.

[0030] As a further solution of the present invention: The acquisition method of the time difference sequence is as follows:

[0031] Use the Kolmogorov - Smirnov test method to determine whether the monitoring period - event number change curve follows a normal distribution;

[0032] If the monitoring period - event number change curve follows a normal distribution, determine the event-intensive sub-period based on the three-standard-deviation criterion;

[0033] If the monitoring period - event number change curve does not follow a normal distribution, perform curve analysis on the monitoring period - event number change curve to determine the event-intensive sub-period;

[0034] Obtain the time when the event is triggered by the company within the monitoring period, and the time of the event-intensive sub-period;

[0035] Perform a difference process on the time when the company triggers an event and the time of the event-intensive sub-period to obtain the event time difference;

[0036] Obtain the event time differences within multiple monitoring periods and construct a time difference sequence.

[0037] As a further solution of the present invention: The method of performing curve analysis to determine the event-intensive sub-period is as follows:

[0038] Obtain multiple peak sections within the monitoring period - event number change curve, and use the monitoring sub-period corresponding to the peak section as the event-intensive sub-period.

[0039] As a further solution of the present invention: The determination method of the same type of preferred ability person is as follows:

[0040] If there is a periodic correlation, from the personnel database, obtain the number of successful processing times and the processing time of the same category of events in the frequent category by the AI electronic work card wearer during the event-intensive sub-cycle.

[0041] Perform numerical analysis on the number of successful processing times and the processing time to determine the processing ability of the AI electronic work card wearer.

[0042] If the processing ability of the AI electronic work card wearer is higher than the preset ability threshold, mark the wearer of the AI electronic work card as an excellent ability person of the same kind.

[0043] As a further solution of the present invention: the acquisition method of the processing ability is as follows:

[0044] Perform a ratio process on the processing time and the time length of the event-intensive sub-cycle to obtain a processing time ratio.

[0045] Perform a ratio process on the number of successful processing times of the same category of events and the number of occurrences of frequent category events during the event-intensive sub-cycle to obtain an event processing ratio.

[0046] Based on the processing time ratio and the event processing ratio, calculate the processing ability of the AI electronic work card wearer through a weighted formula.

[0047] As a further solution of the present invention: the method for predicting the time of the event-intensive sub-cycle and the category of customer events is as follows:

[0048] Obtain the time of the event-intensive sub-cycle from historical data and construct a dense time data group.

[0049] Obtain from historical data the category of customer events caused by company-triggered events and construct an induced category data group.

[0050] Through the long short-term memory network algorithm and the decision tree model algorithm, establish an event category prediction model to predict the time of the event-intensive sub-cycle and the category of customer events, and obtain a predicted sub-cycle and a predicted event category.

[0051] As a further solution of the present invention: an event intelligent detection system for an AI electronic work card includes the following modules:

[0052] Event classification module: used to collect audio and scene pictures, identify events and classify the events, and construct an event classification table.

[0053] Frequency analysis module: based on the event classification table, construct a deviation data set, and used to perform frequency analysis on the category of customer events in the deviation data set to obtain a frequent category.

[0054] Association analysis module: used to quantitatively analyze the cycle time of frequent categories, obtain the time of event-intensive sub-cycles, acquire the time when the company triggers events, and through statistical analysis, determine whether there is a periodic association between the time when the company triggers events and the time of event-intensive sub-cycles;

[0055] Capability selection module: If there is a periodic association, quantitatively analyze the processing capabilities of AI electronic name tag wearers for customer events of frequent categories, determine the preferred capable persons of the same category, and construct a dataset of capable persons;

[0056] Allocation prediction module: Construct an event category prediction model, predict the time of event-intensive sub-cycles and the categories of customer events, and arrange the preferred capable persons of the same category to the predicted event categories in the predicted sub-cycles.

[0057] Advantages of the present invention:

[0058] (1) By collecting audio and scene pictures through AI electronic name tags, fusing multi-dimensional features through algorithms, and matching with preset event classification labels, various customer events can be identified, and an event classification table can be constructed, enabling enterprises to quickly understand the events occurring during the interaction with customers, providing key data support for in-depth insight into customer needs and market dynamics; Conduct frequency analysis on customer event categories, and use chi-square test and Poisson distribution to determine frequent categories; It is beneficial to distinguish between frequently occurring and occasional event categories, and quantitatively analyze the occurrence patterns of different categories of events. Enterprises can, based on these analysis results, reasonably allocate resources such as manpower and material resources, give priority to handling frequent events, improve resource utilization efficiency, and avoid resource waste.

[0059] (2) Quantitatively analyze the cycle time of frequent categories, determine the time of event-intensive sub-cycles, and judge the periodic association between the time when the company triggers events and the time of event-intensive sub-cycles through the autocorrelation function ACF analysis method. Based on the periodic association, enterprises can make preparations in advance, effectively improve the enterprise's control ability over business, and reduce potential risks.

[0060] (3) If there is a periodic association, quantitatively analyze the capabilities of AI electronic name tag wearers in handling customer events of frequent categories, screen out the preferred capable persons of the same category. Enterprises can, based on these ability evaluation results, reasonably arrange personnel, allocate personnel with strong processing capabilities to key event handling scenarios, improve the success rate and efficiency of event handling; Construct an event category prediction model, predict the time of event-intensive sub-cycles and the categories of customer events, enabling enterprises to predict the business development trend in advance, plan resources in advance, and formulate coping strategies. Description of the drawings

[0061] The present invention will be further described below with reference to the accompanying drawings.

[0062] Figure 1 It is a flowchart of an event intelligent detection method for an AI electronic name tag of the present invention;

[0063] Figure 2 It is a module diagram of an event intelligent detection system for an AI electronic name tag in the present invention. Specific embodiments

[0064] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the protection scope of the present invention.

[0065] Embodiment 1

[0066] Please refer to Figure 1 As shown, the present invention is an event intelligent detection method for an AI electronic name tag, including the following steps:

[0067] Step 1: Collect audio and scene pictures based on the AI electronic name tag, match them with event classification tags, identify events and classify the events, and construct an event classification table;

[0068] In some embodiments, the conversation audio between the wearer of the electronic name tag and the customer is collected through the microphone on the AI electronic name tag, and the camera synchronously captures the scene picture;

[0069] After data desensitization and encryption processing of the conversation audio and the scene picture, use the NLP algorithm to extract the keywords of the conversation audio, combine the OpenCV algorithm to identify the body movements and expressions in the scene picture, and fuse multi-dimensional features through the rule engine to automatically match the preset event classification tags;

[0070] Collect the audio and scene pictures within the monitoring period, extract multi-dimensional features. If the features meet the preset event classification tags, mark the audio and scene pictures within the monitoring period as events;

[0071] Based on the preset event tags, set different categories and construct an event classification table;

[0072] Exemplarily, through the wearer's AI electronic name tag, collect the conversation audio with the customer within the monitoring period. Through NLP analysis, extract the keywords "too high cost" and "deception" from the conversation audio. The OpenCV algorithm identifies the customer's action of pointing at the contract and rapid body swings. Use the keywords and actions as multi-dimensional features. Since the features meet the preset event tags ["deception", "action of pointing at the contract"], mark the audio and scene pictures within the monitoring period as events;

[0073] By collecting the audio and video data of AI electronic work badges and customers, collecting customer events and classifying the customer events into different categories, an event classification table is constructed.

[0074] Exemplarily, the event categories include: questioning, consulting, cooperation intention, clarification, and other events.

[0075] Step 2: Statistically classify the customer event categories based on the event classification table, construct a deviation data set, perform frequency analysis on the categories of customer events in the deviation data set, and obtain the frequent categories.

[0076] The construction method of the deviation data set is as follows:

[0077] During the monitoring period, based on the event classification table, obtain customer events of different categories and construct a customer event data set.

[0078] Use the chi-square test based on the customer event data set to calculate the chi-square statistic of different event categories.

[0079] Based on the chi-square statistic, calculate the p-value and determine whether the customer event data of different categories in the customer event data set conforms to the Poisson distribution.

[0080] If it conforms, split the customer event data set into a fitting data set; otherwise, split it into a deviation data set.

[0081] Specifically, calculate the p-value of the customer events of different categories in the customer event data set. If the p-value is less than the significance level α, it is considered that the customer events of this category do not conform to the Poisson distribution.

[0082] If the p-value is greater than or equal to the significance level α, it is considered that the customer events of this category conform to the Poisson distribution.

[0083] Preferably, the significance level α is taken as 0.05.

[0084] Through the Poisson distribution formula: Obtain the Poisson distribution probability P(k) of different category data in the fitting data set. k represents the value of the random variable, that is, the number of occurrences of each event category during the monitoring period. is the parameter of the Poisson distribution, represented by the mean value of the number of occurrences of each event category during the monitoring period, and e is the natural constant.

[0085] It should be noted that the Poisson distribution is used in the analysis of customer event categories to determine whether the occurrence patterns of different categories of customer events conform to the characteristics of a random and independent distribution: through the chi-square test and significance level (such as α = 0.05), the event dataset is split into a "fitting dataset" that conforms to the Poisson distribution and a "deviating dataset" that does not. The former can quantify the probability of an event occurring k times within the monitoring period through the Poisson distribution formula, providing a probability basis for resource allocation for stable and frequent events; the latter analyzes the number of consecutive occurrences and the number of intervals of events to identify high-frequency or sudden concentration events with abnormal distributions, assisting in mining data characteristics and detecting abnormal patterns, thereby providing data support for the frequency analysis of customer events, rational resource allocation, and early warning of potential problems;

[0086] Conduct a frequency analysis on the same customer event categories in the deviating dataset to obtain customer events in the frequent categories of the deviating dataset;

[0087] Among them, the method of conducting a frequency analysis on the same customer event categories in the deviating dataset is as follows:

[0088] Based on the deviating dataset, obtain the number of consecutive occurrences and the number of intervals of the same customer event category within the monitoring period;

[0089] Specifically, divide the monitoring period into multiple sub-periods. If a customer event is identified within a sub-period, and in the time dimension, the same customer event is also identified in adjacent sub-periods, and the event categories are the same, that is, the customer events that occur in the sub-period and adjacent sub-periods are consecutive;

[0090] Through the formula: Obtain the event frequency value Lp, where Lx is the number of consecutive occurrences of the same customer event category within the monitoring period, Lg is the number of intervals of the same customer event category within the monitoring period, Ls is the total number of the same customer event category within the monitoring period, and b1 and b2 take 1.64 and 0.82 respectively;

[0091] Obtain the event frequency values Lp of different categories of customer events in the deviating dataset, and compare the event frequency values with a preset frequency threshold;

[0092] If the event frequency value of different categories of customer events is higher than the preset frequency threshold, then mark the customer event category as a frequent category; otherwise, mark it as an occasional category;

[0093] It should be noted that the functions of determining the frequent categories are as follows: Function 1, locate abnormal high-frequency events and identify business pain points. By calculating the event frequency values, screen out the frequent categories where the actual occurrence frequency is significantly higher than the random expectation from the events deviating from the Poisson distribution, directly locate the "abnormal high-frequency events" in the current business due to concentrated outbreaks, continuous occurrences, or being driven by specific factors, and provide directions for quickly investigating potential problems such as product defects and service vulnerabilities;

[0094] Function 2, dynamically allocate resources and efficiently respond to sudden concentrated demands. The frequent categories correspond to the non-randomly aggregated customer demands in the short term. By marking such events, enterprises can break through the traditional resource allocation mode based on stable distribution. For the sudden concentration of high-frequency events (such as concentrated consultations caused by policy adjustments and intensive complaints caused by system failures), allocate special manpower in real time or activate emergency plans to avoid the low processing efficiency or damaged customer experience caused by resource lag.

[0095] The technical solution of this embodiment is as follows: Based on the AI electronic work badges, collect audio and scene pictures, match them with event classification tags, identify events and classify the events, construct an event classification table, statistically classify the customer event categories based on the event classification table to construct a deviation dataset, and perform frequency analysis on the categories of customer events in the deviation dataset to obtain the frequent categories, enabling enterprises to quickly understand various events that occur during the interaction with customers and providing key data support for understanding market dynamics.

[0096] Embodiment 2

[0097] As Figure 1 shown, an event intelligent detection method for an AI electronic work badge further includes the following steps:

[0098] Step 3, perform quantitative analysis on the cycle time of the frequent categories to obtain the time of the event-intensive sub-cycle, obtain the time when the company triggers events, and determine whether there is a periodic correlation between the time when the company triggers events and the time of the event-intensive sub-cycle through statistical analysis;

[0099] Based on the customer events of the frequent categories, obtain the number of customer events in multiple sub-cycles of the monitoring cycle, and draw a monitoring cycle - event number change curve;

[0100] Use the Kolmogorov-Smirnov test method to determine whether the monitoring cycle - event number change curve follows a normal distribution;

[0101] It should be noted that the Kolmogorov-Smirnov test method determines whether the data follows a normal distribution by comparing the maximum difference Dn between the empirical distribution function Fn(x) of the sample data and the distribution function F(x) of the theoretical normal distribution, where x is the independent variable of the distribution function;

[0102] By the formula: Obtain the maximum difference Dn between the empirical distribution function Fn(x) of the sample data and the distribution function F(x) of the theoretical normal distribution, where the sample data are the coordinates of the period-event change curve in the monitoring period - event number change curve;

[0103] Calculate the value of the maximum difference Dn between the empirical distribution function Fn(x) of the sample data and the distribution function F(x) of the theoretical normal distribution. If > > , it is considered that the monitoring period - event number change curve follows the normal distribution; otherwise, it does not, where α is the significance level value;

[0104] If the monitoring period - event number change curve follows the normal distribution, based on the three - standard - deviation criterion, determine the event - intensive sub - period;

[0105] Specifically, calculate the mean and the standard deviation s of the monitoring period - event number change curve, calculate -3s, +3s], and the monitoring sub - periods within the range of -3s, +3s] are the event - intensive sub - periods;

[0106] If the monitoring period - event number change curve does not follow the normal distribution, obtain multiple peak sections within the monitoring period - event number change curve, and use the monitoring sub - periods corresponding to the peak sections as the event - intensive sub - periods;

[0107] Obtain the time of the events caused by the company and the time of the event - intensive sub - periods within the monitoring period, and perform a difference process on the time of the events caused by the company and the time of the event - intensive sub - periods to obtain the event time difference;

[0108] It should be noted that the events caused by the company include: marketing event of the company, social opinion event related to the company, product or service - related event;

[0109] Among them, the time of the events caused by the company and the time of the event - intensive sub - periods are both the starting times of the events;

[0110] Obtain the event time differences within multiple monitoring periods and construct a time - difference sequence;

[0111] Determine whether there is a periodic correlation between the time of the events caused by the company and the time of the event - intensive sub - periods through the autocorrelation function ACF analysis method;

[0112] By the formula: Obtain the autocorrelation coefficient , where is the lag step, is the mean of the time difference sequence, is the event time difference in the time difference sequence, j is the number of each event time difference in the time difference sequence, and m is the total number of event time differences;

[0113] Based on the autocorrelation coefficient , through the formula: obtain the Q statistic, is the maximum value of the lag step;

[0114] If the Q statistic follows a chi-square distribution with degrees of freedom , and the p-value of the time difference sequence is lower than the significance level α, it is considered that there is a periodic association between the time when the company triggers an event and the time of the event-intensive sub-cycle. Otherwise, it cannot be considered that there is a periodic association between the time when the company triggers an event and the time of the event-intensive sub-cycle;

[0115] It should be noted that the analysis of whether there is a periodic association has the following effects: Effect 1, improving business prediction ability: This periodic association provides a strong basis for enterprises to predict business trends;

[0116] Effect 2, continuously optimizing resource allocation: If there is a periodic association, enterprises can allocate resources in advance according to this rule.

[0117] Step Four: If there is a periodic association, quantitatively analyze the processing ability of AI electronic work badge wearers for frequent types of customer events, determine the preferred ability holders of the same type, and construct a dataset of ability holders; [[ID=3l]]

[0118] If there is a periodic association, obtain AI electronic work badge wearers from the personnel database, analyze the processing ability for frequent types of customer events, and obtain the processing ability Nl;

[0119] Among them, the processing ability includes: the number of successful processing times of the same type of events and the processing time of AI electronic work badge wearers for frequent types during the event-intensive sub-cycle;

[0120] Process the ratio of the processing time to the time length of the event-intensive sub-cycle to obtain the processing time ratio;

[0121] [[ID=,41]]Obtain the number of occurrences of frequent type events during the event-intensive sub-cycle, and process the ratio of the number of successful processing times of the same type of events to the number of occurrences of frequent type events during the event-intensive sub-cycle to obtain the event processing ratio;

[0122] Through the formula: Obtain the processing ability Nl of the wearer of the AI electronic work badge, where c and d are respectively taken as 0.85 and 0.46, Sj is the processing time ratio, and Sc is the event processing ratio;

[0123] Compare the processing ability Nl of the wearer of the AI electronic work badge with a preset ability threshold. If the processing ability Nl of the wearer of the AI electronic work badge is higher than the preset ability threshold, mark the wearer of the AI electronic work badge as a person with preferred ability of the same type; otherwise, do not process.

[0124] Obtain the persons with preferred ability of the same type in different event categories in the frequent category, and construct a dataset of persons with ability.

[0125] It should be noted that the role of determining the persons with preferred ability of the same type is as follows: Role 1, improve the event processing efficiency. In the event-intensive sub-cycle, the persons with preferred ability of the same type can more efficiently and accurately handle the frequent category events with their excellent processing ability, and can also prevent the accumulation and deterioration of problems, which is beneficial to the smooth progress of the business process and improves the overall operation efficiency of the enterprise;

[0126] Role 2, optimize the human resource allocation. Determining the persons with preferred ability of the same type provides a scientific basis for the enterprise to reasonably allocate human resources, improves the utilization efficiency of human resources, and ensures that the enterprise realizes the maximization of benefits under limited human conditions.

[0127] The technical solution of this embodiment is as follows: Quantitatively analyze the cycle time of the frequent category to obtain the time of the event-intensive sub-cycle, obtain the time when the company triggers an event, and through statistical analysis, determine whether there is a periodic correlation between the time when the company triggers an event and the time of the event-intensive sub-cycle; if there is a periodic correlation, quantitatively analyze the processing ability of the wearer of the AI electronic work badge for the customer events of the frequent category, determine the persons with preferred ability of the same type, and screen out the persons with preferred ability of the same type. The enterprise can reasonably arrange personnel according to these ability evaluation results, and allocate the personnel with strong processing ability to key event processing scenarios to improve the success rate and efficiency of event processing.

[0128] Embodiment 3

[0129] As Figure 1 shown, an event intelligent detection method for an AI electronic work badge further includes the following steps:

[0130] Step Five, construct an event category prediction model, predict the time of the event-intensive sub-cycle and the category of customer events to obtain a predicted sub-cycle and a predicted event category, and arrange the persons with preferred ability of the same type to the predicted event category of the predicted sub-cycle;

[0131] Obtain the time of the event-intensive sub-cycle from historical data and construct a dense time data group;

[0132] Obtain from historical data the categories of customer events caused by company-triggered events, and construct an induced category data group;

[0133] Establish an event category prediction model through the long short-term memory network algorithm and the decision tree model algorithm, predict the time of the event-intensive sub-cycle and the categories of customer events, and obtain the predicted sub-cycle and the predicted event categories;

[0134] It should be noted that the dense time data group is predicted through the long short-term memory network algorithm, and the categories of customer events are predicted through the decision tree model algorithm;

[0135] In some embodiments, data preprocessing, feature selection, and model training are performed on the dense time data group and the induced category data group to construct an event category prediction model;

[0136] Input the company-triggered events based on the event category prediction model to obtain the predicted sub-cycle and the predicted event categories;

[0137] Based on the predicted sub-cycle and the predicted event categories, arrange the same type of preferred ability personnel to the predicted event categories of the predicted sub-cycle.

[0138] It should be noted that the functions of determining the predicted sub-cycle and the predicted event categories are as follows: Function 1, advance planning and resource allocation. By determining the predicted sub-cycle and the predicted event categories, the enterprise can predict in advance the business peak period and the time and types of various events, which is conducive to reducing the possibility of chaos in temporary resource allocation and improving the resource utilization efficiency;

[0139] Function 2, optimize personnel arrangement. Precisely arrange the same type of preferred ability personnel to the predicted event categories of the predicted sub-cycle, give full play to their advantages, avoid waste of human resources, achieve the optimal allocation of human resources, and improve the efficiency and quality of event handling;

[0140] Function 3, risk warning and management. Determining the predicted sub-cycle and the predicted event categories helps the enterprise identify potential risks in advance. When risks occur, it can also quickly take countermeasures based on the prior preparations, reduce the losses caused by risks, enhance the enterprise's risk resistance ability, and ensure the stable operation of the enterprise.

[0141] The technical solution of this embodiment is: construct an event category prediction model, predict the time of the event-intensive sub-cycle and the categories of customer events, obtain the predicted sub-cycle and the predicted event categories, arrange the same type of preferred ability personnel to the predicted event categories of the predicted sub-cycle, and predict the time of the event-intensive sub-cycle and the categories of customer events, so that the enterprise can predict in advance the business development trend, plan resources in advance, and formulate countermeasures.

[0142] Embodiment 4

[0143] An event intelligent detection system for an AI electronic work badge further includes the following modules:

[0144] Event classification module: Collect audio and scene images based on the AI electronic work badge, match them with event classification labels, identify events and classify them, and construct an event classification table;

[0145] Frequency analysis module: Construct a deviation data set based on the event classification table to obtain customer event categories, perform frequency analysis on the categories of customer events in the deviation data set, and obtain frequent categories;

[0146] Association analysis module: Used to perform quantitative analysis on the cycle time of frequent categories to obtain the time of event-intensive sub-cycles, obtain the time of events caused by the company, and determine whether there is a periodic association between the time of events caused by the company and the time of event-intensive sub-cycles through statistical analysis;

[0147] Capability selection module: If there is a periodic association, it is used to perform quantitative analysis on the processing capabilities of the AI electronic work badge wearer for customer events of frequent categories, determine the preferred capable persons of the same type, and construct a capable person data set;

[0148] Allocation prediction module: Used to construct an event category prediction model, predict the time of event-intensive sub-cycles and the categories of customer events, obtain prediction sub-cycles and prediction event categories, and arrange the preferred capable persons of the same type to the prediction event categories of the prediction sub-cycles.

[0149] The above has described a specific embodiment of the present invention in detail, but the content described is only the preferred embodiment of the present invention and cannot be considered as limiting the scope of implementation of the present invention. All equivalent changes and improvements made according to the scope of the application of the present invention should still fall within the scope covered by the present invention.

Claims

1. An event intelligent detection method for an AI electronic work card, characterized in that: It includes the following steps: Collect audio and scene pictures, identify events and classify the events to construct an event classification table; Among them, the conversation audio between the wearer of the AI electronic work card and the customer is collected through the microphone on the AI electronic work card, and the camera synchronously captures the scene pictures; Construct a deviation data set based on the event classification table, perform frequency analysis on the categories of customer events in the deviation data set to obtain frequent categories; The determination method of the frequent category is as follows: Obtain customer events of different categories within the monitoring period to construct a customer event data set; Based on the customer event data set, use the chi-square test to judge whether the customer events of different categories in the customer event data set conform to the Poisson distribution; If it conforms, split the customer event data set into a fitting data set, otherwise, split it into a deviation data set; Perform numerical calculations on the categories of customer events in the deviation data set to determine the customer events of the frequent category; Perform quantitative analysis on the cycle time of the frequent category to obtain the time of the event-intensive sub-cycle, obtain the time when the company triggers an event, and through statistical analysis, determine whether there is a periodic association between the time when the company triggers an event and the time of the event-intensive sub-cycle; The judgment method of whether there is a periodic association is as follows: Based on the customer events of the frequent category, obtain the number of customer events in multiple monitoring sub-cycles of the monitoring period, and draw a monitoring period - event number change curve; Perform statistical analysis on the monitoring period - event number change curve to obtain a time difference sequence; Based on the time difference sequence, through the autocorrelation function analysis method and chi-square distribution verification, determine whether there is a periodic association between the time when the company triggers an event and the time of the event-intensive sub-cycle; The acquisition method of the time difference sequence is as follows: Use the Kolmogorov-Smirnov test method to judge whether the monitoring period - event number change curve follows a normal distribution; If the monitoring period - event number change curve follows a normal distribution, based on the 3-fold standard deviation criterion, determine the event-intensive sub-cycle; If the monitoring period - event number change curve does not follow a normal distribution, perform curve analysis on the monitoring period - event number change curve to determine the event-intensive sub-cycle; Obtain the time when the company triggers an event within the monitoring period, and the time of the event-intensive sub-cycle; Perform difference processing on the time when the company triggers an event and the time of the event-intensive sub-cycle to obtain an event time difference; Obtain the event time differences in multiple monitoring periods to construct a time difference sequence; The method of performing curve analysis to determine the event-intensive sub-cycle is as follows: Obtain multiple peak sections in the monitoring period - event number change curve, and use the monitoring sub-cycle corresponding to the peak section as the event-intensive sub-cycle; If there is a periodic association, perform quantitative analysis on the processing ability of the wearer of the AI electronic work card for the customer events of the frequent category to determine the preferred ability persons of the same type; Construct an event category prediction model to predict the time of the event-intensive sub-cycle and the category of customer events, and arrange the preferred ability persons of the same type to the predicted event category of the prediction sub-cycle; The method of predicting the time of the event-intensive sub-cycle and the category of customer events is as follows: Obtain the time of the event-intensive sub-cycle from historical data and construct a dense time data group; Obtain the categories of customer events caused by company-triggered events from historical data and construct an induced category data group; Among them, company-triggered events include: company marketing activity events, social opinion events related to the company, and product or service-related events; Establish an event category prediction model through the long short-term memory network algorithm and the decision tree model algorithm to predict the time of the event-intensive sub-cycle and the category of customer events, and obtain the predicted sub-cycle and the predicted event category.

2. The event intelligent detection method of an AI electronic work card according to claim 1, characterized in that: The method for performing numerical calculations on the deviation data set to determine the customer events of the frequent occurrence category is as follows: Based on the deviation data set, obtain the number of consecutive occurrences and the number of intervals of the same customer event category within the monitoring period; Based on the number of consecutive occurrences and the number of intervals within the monitoring period, obtain the event frequency value through a weighted formula; If the event frequency values of different categories of customer events are higher than the preset frequency threshold, mark the customer event category as the frequent occurrence category.

3. The event intelligent detection method of an AI electronic work card according to claim 1, characterized in that: The determination method of the same-kind preferred ability person is as follows: From the personnel database, obtain the number of successful processing times and the processing time of the same category of events in the frequent occurrence category by the AI electronic work card wearer during the event-intensive sub-cycle; Perform numerical analysis on the number of successful processing times and the processing time to determine the processing ability of the AI electronic work card wearer; If the processing ability of the AI electronic work card wearer is higher than the preset ability threshold, mark the wearer of the AI electronic work card as the same-kind preferred ability person.

4. The event intelligent detection method of an AI electronic work card according to claim 1, characterized in that: The acquisition method of the processing ability is as follows: Perform a ratio process on the processing time and the time length of the event-intensive sub-cycle to obtain a processing time ratio; Perform a ratio process on the number of successful processing times of the same category of events and the number of occurrences of events in the frequent occurrence category during the event-intensive sub-cycle to obtain an event processing ratio; Based on the processing time ratio and the event processing ratio, calculate the processing ability of the AI electronic work card wearer through a weighted formula.

5. An event intelligent detection system for an AI electronic work badge, which is used to implement the event intelligent detection method for an AI electronic work badge described in any one of claims 1-4, and is characterized in that: It includes the following modules: Event classification module: used to collect audio and scene pictures, identify events and classify events, and construct an event classification table; Among them, collect the dialogue audio between the electronic work card wearer and the customer through the microphone on the AI electronic work card, and the camera synchronously captures the scene pictures; Frequency analysis module: construct a deviation data set based on the event classification table, used to perform frequency analysis on the categories of customer events in the deviation data set to obtain the frequent occurrence category; The determination method of the frequent occurrence category is as follows: Obtain different categories of customer events within the monitoring period and construct a customer event data set; Based on the customer event data set, use the chi-square test to judge whether the different categories of customer events in the customer event data set conform to the Poisson distribution; If they conform, split the customer event data set into a fitting data set, otherwise, split it into a deviation data set; Perform numerical calculations on the customer event categories in the deviation data set to determine the customer events of the frequent occurrence category; Association analysis module: used to quantitatively analyze the cycle time of frequent categories, obtain the time of event-intensive sub-cycles, acquire the time when the company triggers events, and determine whether there is a periodic association between the time when the company triggers events and the time of event-intensive sub-cycles through statistical analysis; The judgment method for whether there is a periodic association is as follows: Based on customer events of frequent categories, obtain the number of customer events in multiple monitoring sub-cycles of the monitoring cycle, and draw a monitoring cycle - event number change curve; Perform statistical analysis on the monitoring cycle - event number change curve to obtain a time difference sequence; Based on the time difference sequence, determine whether there is a periodic association between the time when the company triggers events and the time of event-intensive sub-cycles through autocorrelation function analysis method and chi-square distribution test; The acquisition method of the time difference sequence is as follows: Use the Kolmogorov-Smirnov test method to determine whether the monitoring cycle - event number change curve follows a normal distribution; If the monitoring cycle - event number change curve follows a normal distribution, determine the event-intensive sub-cycle based on the 3-fold standard deviation criterion; If the monitoring cycle - event number change curve does not follow a normal distribution, perform curve analysis on the monitoring cycle - event number change curve to determine the event-intensive sub-cycle; Acquire the time when the company triggers events within the monitoring cycle and the time of the event-intensive sub-cycle; Perform a difference process on the time when the company triggers events and the time of the event-intensive sub-cycle to obtain an event time difference; Acquire the event time differences within multiple monitoring cycles and construct a time difference sequence; The method for performing curve analysis to determine the event-intensive sub-cycle is as follows: Acquire multiple peak sections within the monitoring cycle - event number change curve, and use the monitoring sub-cycles corresponding to the peak sections as event-intensive sub-cycles; Capability selection module: If there is a periodic association, quantitatively analyze the processing capabilities of AI electronic name tag wearers for customer events of frequent categories, determine the preferred ability persons of the same category, and construct a dataset of ability persons; Allocation prediction module: Construct an event category prediction model to predict the time of event-intensive sub-cycles and the category of customer events, and arrange the preferred ability persons of the same category to the predicted event category of the prediction sub-cycle; The method for predicting the time of event-intensive sub-cycles and the category of customer events is as follows: Obtain the time of event-intensive sub-cycles from historical data and construct a dense time data group; Obtain the categories of customer events caused by the company's triggering events from historical data and construct an induced category data group; Among them, the company's triggering events include: the company's marketing activity events, social opinion events related to the company, and product or service related events; Through the long short-term memory network algorithm and decision tree model algorithm, establish an event category prediction model to predict the time of event-intensive sub-cycles and the category of customer events, and obtain a prediction sub-cycle and a predicted event category.

Images