AI-based enterprise customer operation and maintenance demand prediction management system and method

Abstract

The application relates to the technical field of customer operation and maintenance demand management, in particular to an enterprise customer operation and maintenance demand prediction management system and method based on AI, which discloses an enterprise customer operation and maintenance prediction demand classification module, a customer set sequence generation module and a customer operation and maintenance intervention guidance module. The system uses a feature fusion method and a time sequence model to regularly predict the operation and maintenance correlation features of each customer of an enterprise, to perform set classification on customers with similar operation and maintenance demands, to build unified operation and maintenance demand prediction models for various customers, to perform deep operation and maintenance demand simulation analysis on various customer sets, to improve the operation and maintenance demand analysis efficiency and precision of each customer, to intervene in the operation and maintenance management of various customer sets according to the operation and maintenance demand sequence, and to accurately and efficiently lock the customer that needs to be first intervened in the operation and maintenance management through deep analysis of the operation and maintenance demands of each customer in the customer set.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of customer operation and maintenance demand management, in particular to an AI-based enterprise customer operation and maintenance demand prediction management system and method. BACKGROUND

[0002] With the deepening of enterprise digital transformation, the scale and complexity of infrastructure such as IT equipment, industrial controllers and network systems grow exponentially, and customer operation and maintenance demand management faces severe challenges. The traditional operation and maintenance mode relies on manual experience and regular inspection, and it is difficult to cope with dynamic scenarios such as equipment performance degradation, business load fluctuation and multi-device coupling failure, resulting in high operation and maintenance costs, low response efficiency and insufficient fault root cause analysis. In recent years, data-driven intelligent operation and maintenance technology has gradually emerged, which realizes fault prediction and resource scheduling by collecting equipment operation data, but the existing solutions generally have the following technical bottlenecks:

[0003] Single customer classification method: only based on static attributes (such as industry, scale) classification, without combining dynamic operation and maintenance demand characteristics (such as fault frequency, equipment aging degree, business peak period, etc.) for real-time clustering, resulting in that the classification result cannot reflect the real demand difference.

[0004] Lack of in-depth analysis of operation and maintenance decision: the existing system stays in the demand prediction stage, and does not form a closed loop with operation and maintenance resource scheduling and intervention strategy optimization, and cannot realize the whole process intelligentization of "accurate prediction - efficient configuration - root cause management" through demand simulation analysis.

[0005] The existing technology has problems such as one-sided feature analysis, inefficient resource allocation and extensive strategy formulation in customer operation and maintenance demand management, and it is difficult to cope with complex device clusters and dynamic business scenarios. Therefore, the application provides an AI-based enterprise customer operation and maintenance demand prediction management system and method. SUMMARY

[0006] In view of the deficiencies of the prior art, the purpose of the application is to provide an AI-based enterprise customer operation and maintenance demand prediction management system and method.

[0007] To achieve the above purpose, the application provides the following technical scheme:

[0008] The AI-based enterprise customer operation and maintenance demand prediction management system comprises an enterprise customer operation and maintenance prediction demand classification module, a customer set sequence generation module and a customer operation and maintenance intervention guidance module.

[0009] The enterprise customer operation and maintenance prediction demand classification module is used to set the customer operation and maintenance demand prediction period of the enterprise, and the enterprise determines the three-dimensional operation and maintenance feature prediction set of each customer based on the customer operation and maintenance demand prediction period, and further generates a plurality of operation and maintenance demand customer sets.

[0010] The customer set sequence generation module controls each operation and maintenance demand customer set to generate various types of operation and maintenance correlation prediction fusion features, builds an operation and maintenance demand unified prediction model of each operation and maintenance demand customer set, and then determines an operation and maintenance demand comprehensive prediction index of each operation and maintenance demand customer set to generate a customer set sequence.

[0011] The customer operation and maintenance intervention guidance module first determines a simulation operation and maintenance recommendation strategy of the operation and maintenance demand customer set in the first place, obtains operation and maintenance root cause residual values of each customer in the operation and maintenance demand customer set, sorts each customer from large to small according to the numerical values of the operation and maintenance root cause residual values, and intervenes in the operation and maintenance management of each customer in sequence according to the sorting. When all the customers in the operation and maintenance demand customer set complete the operation and maintenance management, a simulation operation and maintenance recommendation strategy of the operation and maintenance demand customer set in the next place is determined, and the cycle is repeated.

[0012] Further, the periodic determination method of the three-dimensional operation and maintenance feature prediction set of the customer is as follows: whenever a customer operation and maintenance demand prediction cycle ends, various types of operation and maintenance correlation features of the customer are collected, the various types of operation and maintenance correlation features are combined into an operation and maintenance correlation feature set in the form of a feature set, and operation and maintenance correlation feature sets corresponding to the previous continuous c customer operation and maintenance demand prediction cycles are collected. The c+1 operation and maintenance correlation feature sets are combined into an operation and maintenance correlation time sequence set in the form of a time sequence, and the operation and maintenance correlation time sequence set is imported into the operation and maintenance correlation feature prediction model. The operation and maintenance correlation feature prediction model outputs a three-dimensional operation and maintenance feature prediction set.

[0013] Further, the generation method of a type of operation and maintenance correlation prediction fusion feature of an operation and maintenance demand customer set is as follows: all customers included in an operation and maintenance demand customer set are determined, a three-dimensional operation and maintenance feature prediction set of each customer is obtained, a type of operation and maintenance correlation feature is selected, the type of operation and maintenance correlation feature in each three-dimensional operation and maintenance feature prediction set is extracted, all the extracted type of operation and maintenance correlation features are fused, and a type of operation and maintenance correlation prediction fusion feature is generated.

[0014] Further, the determination method of the operation and maintenance demand comprehensive prediction index of the operation and maintenance demand customer set is as follows: the operation and maintenance demand unified prediction model of an operation and maintenance demand customer set is controlled to simulate for T mi time, and after the T mi time ends, the operation and maintenance demand unified prediction model outputs various types of core physical quantities of the equipment. The various types of core physical quantities are feature extracted, various types of core physical features are extracted, the various types of core physical features are combined into a comprehensive core physical feature set in the form of a feature set, the comprehensive core physical feature set is imported into the operation and maintenance demand model, and the operation and maintenance demand model outputs an operation and maintenance demand comprehensive prediction index.

[0015] Further, the generation method of the customer set sequence: the operation and maintenance demand comprehensive prediction index value is sorted from large to small for each operation and maintenance demand customer set, and the customer set sequence is generated according to the sorting.

[0016] Further, the determination method of the simulation operation and maintenance recommendation strategy of the operation and maintenance demand customer set: the operation and maintenance demand unified prediction model corresponding to the operation and maintenance demand customer set is obtained to perform multiple simulation operations and maintenances, and the operation and maintenance strategy corresponding to the simulation operation and maintenance with the minimum operation and maintenance demand comprehensive prediction index is marked as the simulation operation and maintenance recommendation strategy.

[0017] Further, the acquisition method of the operation and maintenance root cause residual value of the customer: a customer is selected, the simulation operation and maintenance recommendation strategy of the operation and maintenance demand customer set corresponding to the customer in the last period is obtained, the simulation operation and maintenance recommendation strategy in the current period and the simulation operation and maintenance recommendation strategy in the last period are respectively subjected to feature extraction, the simulation operation and maintenance strategy features in the current period and the simulation operation and maintenance strategy features in the last period are extracted, the simulation operation and maintenance strategy features in the current period and the simulation operation and maintenance strategy features in the last period are matched into a simulation operation and maintenance feature group, the simulation operation and maintenance feature group is imported into the operation and maintenance strategy comparison model, and the operation and maintenance strategy comparison model outputs an operation and maintenance strategy residual index AD(s). The operation and maintenance demand comprehensive prediction threshold index is set, when the operation and maintenance demand comprehensive prediction index of the operation and maintenance demand prediction period of the customer in the operation and maintenance demand customer set is greater than or equal to the operation and maintenance demand comprehensive prediction threshold index, the corresponding customer operation and maintenance demand prediction period is marked as an urgent demand period, the urgent demand times is increased by one, and all the urgent demand periods are compared with each other, when the compared two urgent demand periods are adjacent customer operation and maintenance demand prediction periods, the urgent demand duration times is increased by one, finally, the urgent demand times is marked as KS(c), the urgent demand duration times is marked as RY(u), and the actual operation and maintenance possible demand value EW(a) of the customer is calculated through EW(a) = AD(s)*[KS(c)+RY(u)].

[0018] Further, the AI-based enterprise customer operation and maintenance demand prediction management method has the following steps:

[0019] Step one: periodically determine the three-dimensional operation and maintenance feature prediction set of each customer of the enterprise, and further generate multiple operation and maintenance demand customer sets;

[0020] Step two: control each operation and maintenance demand customer set to generate various types of operation and maintenance associated prediction fusion features, and build an operation and maintenance demand unified prediction model for each operation and maintenance demand customer set;

[0021] Step three: determine the operation and maintenance demand comprehensive prediction index of each operation and maintenance demand customer set, and generate a customer set sequence;

[0022] Step four: determine the simulation operation and maintenance recommendation strategy of the operation and maintenance demand customer set ranked first, and obtain the operation and maintenance root cause residual value of each customer in the operation and maintenance demand customer set;

[0023] Step five: sort each customer from large to small according to the numerical value of the operation and maintenance root cause residual value, and intervene in the operation and maintenance management of each customer according to the order.

[0024] Compared with the prior art, the present application has the following beneficial effects:

[0025] The system of the present application uses feature fusion means and a time sequence model to periodically predict the operation and maintenance correlation features of each customer of the enterprise, classifies customers with similar operation and maintenance needs in a set, builds a unified operation and maintenance demand prediction model for various customers, and performs in-depth operation and maintenance demand simulation analysis on each customer set, thereby further improving the operation and maintenance demand analysis efficiency and accuracy of each customer, and intervening in the operation and maintenance management of each customer set in order of operation and maintenance demand, and accurately and efficiently locking the customer that needs to be intervened in operation and maintenance management first through in-depth analysis of the operation and maintenance demand of each customer in the customer set. The system significantly improves the intelligent level of enterprise operation and maintenance, and is especially suitable for industries with large equipment cluster scale and complex business scenarios, and provides customers with "accurate prediction - efficient configuration - root cause management" comprehensive protection. BRIEF DESCRIPTION OF DRAWINGS

[0026] Fig. 1 The figure is the principle block diagram of the system of the present application;

[0027] Fig. 2 The figure is the flow chart of determining the operation and maintenance demand comprehensive prediction index of the operation and maintenance demand customer set;

[0028] Fig. 3 The figure is the flow chart of the method of the present application. DETAILED DESCRIPTION

[0029] Embodiment one, refer to Figs. 1-2 The AI-based enterprise customer operation and maintenance demand prediction management system comprises an enterprise customer operation and maintenance prediction demand classification module, a customer set sequence generation module, and a customer operation and maintenance intervention guidance module.

[0030] The enterprise customer operation and maintenance prediction demand classification module sets the customer operation and maintenance demand prediction period of the enterprise (the period length of the customer operation and maintenance demand prediction period is comprehensively set according to the business characteristics, equipment demand, and other attributes of the customer), and the enterprise determines the three-dimensional operation and maintenance feature prediction set of each customer based on the customer operation and maintenance demand prediction period, and further generates a plurality of operation and maintenance demand customer sets.

[0031] The periodic determination method of the stereoscopic operation and maintenance feature prediction set of the customer is as follows: whenever a customer operation and maintenance demand prediction cycle is completed, the operation and maintenance associated features of the customer (the types of operation and maintenance associated features include device basic attribute features, device running state features, fault and maintenance features, etc., and the above features are the core data directly supporting operation and maintenance demand analysis, but their influence on operation and maintenance demand includes both direct association and indirect driving) are collected, the operation and maintenance associated features are combined into an operation and maintenance associated feature set (the operation and maintenance associated feature set can be used for model training) in the form of a feature set, the operation and maintenance associated feature sets corresponding to the previous c consecutive customer operation and maintenance demand prediction cycles of the customer are collected, the c+1 operation and maintenance associated feature sets are combined into an operation and maintenance associated time sequence set in the form of a time sequence, the operation and maintenance associated time sequence set is input into the operation and maintenance associated feature prediction model to which the customer belongs, and the operation and maintenance associated feature prediction model outputs a stereoscopic operation and maintenance feature prediction set (the output stereoscopic operation and maintenance feature prediction set is the stereoscopic operation and maintenance feature prediction set of the customer).

[0032] Each customer belongs to an operation and maintenance associated feature prediction model, and each operation and maintenance associated feature prediction model is built based on the training of an LSTM model. In the specific implementation, the building process of the operation and maintenance associated feature prediction model of customer A is disclosed: an LSTM model is built, multiple operation and maintenance associated time sequence sets of customer A are collected, the operation and maintenance associated time sequence sets are used as basic data, the built LSTM model is trained, in this process, each operation and maintenance associated time sequence set is given a stereoscopic operation and maintenance feature prediction set (the dimensions of the operation and maintenance associated time sequence sets are different, and direct input into the LSTM may cause gradient disappearance / explosion, so each type of operation and maintenance associated feature in the operation and maintenance associated time sequence set is normalized separately), the stereoscopic operation and maintenance feature prediction set contains each type of operation and maintenance associated feature of customer A in the next customer operation and maintenance demand prediction cycle, then the multiple operation and maintenance associated time sequence sets are divided into a training set, a validation set and a test set according to a specific ratio, and the specific division ratio is determined as 70%:20%:10%. First, the training set is used to repeatedly train the LSTM model, in the training process, the performance of the model in the training stage is verified in time with the help of the validation set, the parameters of the model are adjusted in time according to the verification result, the structure of the model is optimized, and the model is developed in a more accurate and stable direction. Finally, the operation and maintenance associated feature prediction model of customer A is built.

[0033] A set of operation and maintenance demand customers includes multiple customers, and the operation and maintenance demand similarity between each two of the customers is greater than or equal to a set threshold; the operation and maintenance demand similarity between two customers: the two customers are marked as customer M and customer N, each type of operation and maintenance associated feature in the stereoscopic operation and maintenance feature prediction set of customer M is vectorized and converted into a vector M=(M1, M2,..., M n), vectorize each type of operation and maintenance associated feature in the stereoscopic operation and maintenance feature prediction set of customer N, and convert it into vector N=(N1, N2,..., N n ), use the similarity formula to calculate the operation and maintenance requirement similarity between customer M and customer N.

[0034] A customer set sequence generation module generates each type of operation and maintenance associated prediction fusion feature for each operation and maintenance requirement customer set, builds an operation and maintenance requirement unified prediction model for each operation and maintenance requirement customer set, and then determines an operation and maintenance requirement comprehensive prediction index for each operation and maintenance requirement customer set to generate a customer set sequence.

[0035] The generation method of one type of operation and maintenance associated prediction fusion feature of one operation and maintenance requirement customer set: determine all customers included in one operation and maintenance requirement customer set, obtain the stereoscopic operation and maintenance feature prediction set of each customer, select one type of operation and maintenance associated feature, extract the type of operation and maintenance associated feature in each stereoscopic operation and maintenance feature prediction set, fuse all the type of operation and maintenance associated features extracted to generate the type of operation and maintenance associated prediction fusion feature.

[0036] Taking the device running state feature as an example, the feature fusion processing includes but is not limited to the following methods: standardization processing: the device running state feature is usually a numerical data changing with time, such as CPU utilization rate, memory usage rate, disk I / O, etc. of the device. Before fusion, these data need to be standardized to eliminate the influence of different features. The commonly used standardization method is Z-score standardization, the formula is as follows: Wherein, x is the original feature value, μ is the mean of the feature, σ is the standard deviation of the feature; time series fusion: since the device running state feature is time series data, time series analysis method can be used for fusion. For example, sliding average method can be used to smooth the device running state time series of multiple customers, and then weighted average is performed. Assuming that there are n customers' device running state time series {x i1 ,x i2 ,…,x it}(i=1, 2,..., n), for each time point t, the fused feature value X t can be calculated by the following formula: Wherein, wi is the weight of each customer.

[0037] For example, the feature fusion processing of the fault and maintenance feature includes but is not limited to the following ways: fault frequency statistics: the fault and maintenance feature usually includes the fault times, fault types, maintenance time and other information of the equipment, for the fault times and fault types and other information, a statistical method can be used for fusion, for example, the fault times of each customer within a certain time are counted, and then the average fault times of all customers are calculated as the fused fault frequency feature.

[0038] Fault severity evaluation: for the fault type, the fault can be classified and quantified according to the severity of the fault. For example, the fault is divided into slight fault, moderate fault and serious fault, and different weights are given respectively, then the fault severity score of each customer is calculated, and finally the weighted average of the scores of all customers is obtained to get the fused fault severity feature.

[0039] Maintenance time series fusion: for the maintenance time and other time series data, a similar time series fusion method can be used for processing as the device running state feature. For example, the moving average method is used to smooth the maintenance time series of multiple customers, and then the weighted average is performed.

[0040] The building method of the unified operation and maintenance demand prediction model of the operation and maintenance demand customer set: obtain each type of operation and maintenance related prediction fusion feature corresponding to an operation and maintenance demand customer set, import each type of operation and maintenance related prediction fusion feature into a general finite element software, map each type of operation and maintenance related prediction fusion feature to the mapping parameters required for finite element simulation (for example, the device basic attribute feature includes CPU core number, fan air volume and other features, the corresponding mapping parameters include heat source number, thermal conductivity, convective heat transfer coefficient, etc., for example, the device running state feature includes CPU utilization, memory temperature, fan speed, the corresponding mapping parameters include power consumption (W), initial temperature (℃), speed (RPM), for example, the fault and maintenance feature includes historical hard disk fault times, maintenance time and fan model, the corresponding mapping parameters include contact thermal resistance (increased by 10% after hard disk failure), fan efficiency attenuation coefficient), construct the geometric model in the general finite element software, for example, taking a server case as an example, first, parameterized design: fin thickness t, fan position (x, y, z), CPU layout n as adjustable parameters corresponding to the device basic attribute feature, then define the material and boundary conditions, such as the material of the fin is aluminum alloy, the thermal conductivity is defined as 180 W / (m·K), the density is 2700 kg / m 3 such as ambient temperature: 25℃, fan air volume: Q = fan speed x 0.01 m 3 / min (linear mapping of operating state characteristics), perform multi-physics field settings (such as thermal-structural coupling, thermal analysis: calculate CPU power consumption (P = CPU utilization × 200W) as a heat source, simulate temperature distribution), and finally build a unified prediction model of operation and maintenance requirements for the customer set through model verification and validation. The unified prediction model of operation and maintenance requirements can simulate the operation of equipment.

[0041] The customer set sequence is generated by sorting each customer set for maintenance needs in descending order of the comprehensive prediction index of maintenance needs, and then generating the customer set sequence based on the sorting.

[0042] The method for determining the comprehensive forecast index of maintenance needs for a set of maintenance demand customers: Controlling a unified forecasting model for maintenance needs of a set of maintenance demand customers to perform T... mi The duration of the simulation (which is essentially the T-scaling of equipment in the unified prediction model of operation and maintenance needs) mi (Duration of operation), T mi After the duration ends, the unified prediction model for operation and maintenance requirements outputs various types of core physical quantities of the equipment (the types of core physical quantities include temperature distribution cloud maps, airflow velocity vector maps, and thermal resistance / power consumption curves). Feature extraction is performed on each type of core physical quantity (feature extraction methods for temperature distribution cloud maps include basic statistical feature extraction and image feature extraction, feature extraction methods for airflow velocity vector maps include vector field statistical feature extraction, critical path feature extraction, and fluid dynamics parameter feature extraction, and feature extraction methods for thermal resistance / power consumption curves include curve shape features, etc.). The core physical features of each type are extracted and combined into a comprehensive core physical feature set. The comprehensive core physical feature set is imported into the operation and maintenance requirement model, and the operation and maintenance requirement model derives a comprehensive prediction index for operation and maintenance requirements.

[0043] The operation and maintenance demand model is built in the following manner: a deep learning model is built, multiple comprehensive core physical feature sets are collected, the comprehensive core physical feature sets are taken as basic data, the built deep learning model is trained, in this process, an operation and maintenance demand comprehensive prediction index is given to each comprehensive core physical feature set, the value range of the operation and maintenance demand comprehensive prediction index is set to be between 1 and 100, the size of the operation and maintenance demand comprehensive prediction index has a clear meaning, the larger the value is, the greater the operation and maintenance demand of the comprehensive core physical feature set is, that is, the higher the comprehensive risk of the equipment in thermal stability, heat dissipation efficiency, load bearing capacity and the like is, and the more urgent the operation and maintenance demand (such as inspection, maintenance, upgrade and the like) is, then the multiple comprehensive core physical feature sets are divided into a training set, a verification set and a test set according to a specific proportion, the specific division proportion is determined to be 70%:20%:10%, the training set is first used to repeatedly train the deep learning model, in the training process, the performance of the model in the training stage is verified in time by means of the verification set, the parameters of the model are adjusted in time according to the verification result, the structure of the model is optimized, and the model is developed in a more accurate and stable direction, and finally the operation and maintenance demand model is built.

[0044] The customer operation and maintenance intervention guidance module first determines the simulation operation and maintenance recommended strategy of the operation and maintenance demand customer set in the first place, obtains the operation and maintenance root cause residual value of each customer in the operation and maintenance demand customer set, sorts each customer from large to small according to the value of the operation and maintenance root cause residual value, and intervenes in the operation and maintenance management (intervenes in the actual operation and maintenance management) according to the sorting, when all the customers in the operation and maintenance demand customer set complete the operation and maintenance management, the simulation operation and maintenance recommended strategy of the operation and maintenance demand customer set in the next place is determined, the operation and maintenance root cause residual value of each customer in the operation and maintenance demand customer set is obtained, each customer is sorted from large to small according to the value of the operation and maintenance root cause residual value, and each customer is intervened in the operation and maintenance management according to the sorting (that is, the processing process of the previous operation and maintenance demand customer set is repeated in a loop).

[0045] The operation and maintenance demand customer set is determined in the following manner: the operation and maintenance demand unified prediction model corresponding to the operation and maintenance demand customer set is simulated multiple times (each simulation operation and maintenance automatically generates an operation and maintenance strategy, simulates the operation and maintenance processing of the equipment in the operation and maintenance demand unified prediction model, and the operation and maintenance strategies corresponding to each simulation operation and maintenance are different), and the operation and maintenance strategy corresponding to the simulation operation and maintenance with the smallest operation and maintenance demand comprehensive prediction index is marked as the simulation operation and maintenance recommended strategy.

[0046] The method for obtaining the operation and maintenance root cause residual value of the customer: select a customer, obtain the simulation operation and maintenance recommendation strategy of the customer set corresponding to the operation and maintenance demand of the last period, respectively extract the simulation operation and maintenance recommendation strategy of the current period and the simulation operation and maintenance recommendation strategy of the last period, extract the simulation operation and maintenance strategy features of the current period and the simulation operation and maintenance strategy features of the last period, match the simulation operation and maintenance strategy features of the current period and the simulation operation and maintenance strategy features of the last period into a simulation operation and maintenance feature group, import the simulation operation and maintenance feature group into the operation and maintenance strategy comparison model, and the operation and maintenance strategy comparison model exports an operation and maintenance strategy residual index AD(s). Set an operation and maintenance demand comprehensive prediction threshold index (according to the system setting), when the operation and maintenance demand comprehensive prediction index of the customer operation and maintenance demand prediction period in the operation and maintenance demand customer set is greater than or equal to the operation and maintenance demand comprehensive prediction threshold index, mark the corresponding customer operation and maintenance demand prediction period as an urgent demand period, and increase the urgent demand number by one. When the operation and maintenance demand comprehensive prediction index of the customer operation and maintenance demand prediction period in the operation and maintenance demand customer set is less than the operation and maintenance demand comprehensive prediction threshold index, do not mark, compare all urgent demand periods in pairs, when the two compared urgent demand periods are adjacent customer operation and maintenance demand prediction periods, increase the urgent demand duration number by one, when the two compared urgent demand periods are not adjacent customer operation and maintenance demand prediction periods, do not need to increase the urgent demand duration number by one, finally, mark the urgent demand number as KS(c), mark the urgent demand duration number as RY(u), and calculate the actual operation and maintenance possible demand value EW(a) of the customer through EW(a) = AD(s)*[KS(c)+RY(u)].

[0047] The building mode of the operation and maintenance strategy comparison model is as follows: a deep learning model is built, a plurality of simulation operation and maintenance feature groups are collected, each simulation operation and maintenance feature group comprises two simulation operation and maintenance strategy features, the simulation operation and maintenance feature group is taken as basic data, the built deep learning model is trained, in the process, an operation and maintenance strategy residual index is given to each simulation operation and maintenance feature group, the value range of the operation and maintenance strategy residual index is set to 1-5, the size of the operation and maintenance strategy residual index has a clear meaning, the greater the value is, the closer the two simulation operation and maintenance strategy features in the simulation operation and maintenance feature group are, that is, the more consistent the strategies used in the two simulation operations are, that is, the more consistent the simulation operation and maintenance is, and it is extremely possible that the operation and maintenance problem is not solved fundamentally, then the plurality of simulation operation and maintenance feature groups are divided into a training set, a validation set and a test set according to a specific proportion, the specific division proportion is determined as 70%:15%:15%, the training set is used to repeatedly train the deep learning model, in the training process, the performance of the model in the training stage is verified in time by means of the validation set, the parameters of the model are adjusted in time according to the verification result, the structure of the model is optimized, and the model is developed in a more accurate and stable direction, and finally, the operation and maintenance strategy comparison model is built.

[0048] The system of the application uses feature fusion means and a time sequence model, regularly predicts operation and maintenance related features of each customer of an enterprise, classifies customers with similar operation and maintenance needs in a set, builds unified operation and maintenance demand prediction models for various customers, and deeply analyzes the operation and maintenance needs of each customer set, thereby further improving the operation and maintenance demand analysis efficiency and precision of each customer, and sequentially intervening in the operation and maintenance management of each customer set according to the operation and maintenance demand sequence, and accurately and efficiently locking the customer that needs to be first intervened in the operation and maintenance management through deep analysis of the operation and maintenance needs of each customer in the customer set. The system significantly improves the intelligent level of enterprise operation and maintenance, and is especially suitable for industries with large device cluster scale and complex business scenarios (such as cloud computing, intelligent manufacturing, and financial technology), and provides customers with "accurate prediction - efficient configuration - root cause management" comprehensive protection.

[0049] Embodiment two, refer to Fig. 3 The AI-based enterprise customer operation and maintenance demand prediction management method comprises the following steps:

[0050] Step one: regularly determine the three-dimensional operation and maintenance feature prediction set of each customer of an enterprise, and further generate a plurality of operation and maintenance demand customer sets.

[0051] Step two: control each operation and maintenance demand customer set to generate various types of operation and maintenance related prediction fusion features, and build an operation and maintenance demand unified prediction model for each operation and maintenance demand customer set.

[0052] Step three: determine the operation and maintenance demand comprehensive prediction index of each operation and maintenance demand customer set, and generate a customer set sequence.

[0053] Step four: determine the simulation operation and maintenance recommendation strategy of the operation and maintenance demand customer set ranked first, and obtain the operation and maintenance root cause residual value of each customer in the operation and maintenance demand customer set.

[0054] Step five: sort each customer from large to small according to the value of the operation and maintenance root cause residual value, and intervene in the operation and maintenance management of each customer in sequence according to the sorting.

[0055] The above formulas are all dimensionless and the numerical values are calculated. The preset parameters in the formulas are set by a person skilled in the art according to the actual situation.

[0056] The above embodiments can be realized wholly or partially by software, hardware, firmware or any other combination. When realized by software, the above embodiments can be realized wholly or partially in the form of 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, the processes or functions described in the embodiments of the present application are wholly or partially generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network or other programmable devices. The computer instructions can be stored in a computer readable storage medium or transferred from one computer readable storage medium to another, for example, the computer instructions can be transferred from one website, computer, server or data center to another website, computer, server or data center through wired or wireless (such as infrared, wireless, microwave, etc.) mode. The computer readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server, data center and the like containing one or more available medium sets. The available medium can be a magnetic medium (such as a floppy disk, a hard disk, a magnetic tape), an optical medium (such as a DVD) or a semiconductor medium. The semiconductor medium can be a solid state disk.

[0057] It should be understood that in various embodiments of the present application, the size of the serial number of each process does not mean the order of execution, and 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 the present application.

[0058] Those skilled in the art can clearly understand that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be realized by electronic hardware or a combination of computer software and electronic hardware. Whether the functions are performed 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 the present application.

[0059] Those skilled in the art can clearly understand that, for the convenience and brevity of the description, the specific working processes of the above-described system, device and unit can refer to the corresponding processes in the foregoing method embodiments, which will not be repeated here.

[0060] In several embodiments provided in the present application, it should be understood that the disclosed system, device and method can be implemented in other ways. For example, the device embodiments described above are only schematic, for example, the division of the units is only a logical function division, and actual implementation can have another division manner, for example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the units shown or discussed can be indirect coupling or communication connection through some interface, device or unit, and can be electrical, mechanical or other forms.

[0061] If the functions are realized in the form of software function units and sold or used as independent products, they can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application essentially or the parts that make contributions to the prior art or parts of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present application. The foregoing storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various media that can store program codes.

[0062] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art can easily think of changes or replacements within the technical scope disclosed in the present application, which should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. An AI-based enterprise customer operation and maintenance demand prediction management system, characterized in that, The enterprise customer operation and maintenance prediction demand classification module, the customer set sequence generation module, and the customer operation and maintenance intervention guidance module are included. The enterprise customer operation and maintenance prediction demand classification module is used to set the customer operation and maintenance demand prediction period of an enterprise, and the enterprise determines the stereoscopic operation and maintenance feature prediction set for each customer based on the customer operation and maintenance demand prediction period and further generates a plurality of operation and maintenance demand customer sets. The stereoscopic operation and maintenance feature prediction set of a customer is determined in a periodic manner: whenever a customer operation and maintenance demand prediction period ends, the types of operation and maintenance associated features of the customer are collected, including device basic attribute features, device running state features, and fault and maintenance features. The operation and maintenance associated features are combined into an operation and maintenance associated feature set in the form of a feature set, and the operation and maintenance associated feature sets corresponding to the previous c consecutive customer operation and maintenance demand prediction periods of the customer are collected. The c+1 operation and maintenance associated feature sets are combined into an operation and maintenance associated time sequence set in the form of a time sequence, and the operation and maintenance associated time sequence set is input into the operation and maintenance associated feature prediction model to which the customer belongs. The operation and maintenance associated feature prediction model outputs a stereoscopic operation and maintenance feature prediction set. The customer set sequence generation module controls each operation and maintenance demand customer set to generate various types of operation and maintenance associated prediction fusion features, builds an operation and maintenance demand unified prediction model for each operation and maintenance demand customer set, and further determines an operation and maintenance demand comprehensive prediction index for each operation and maintenance demand customer set to generate a customer set sequence. The customer operation and maintenance intervention guidance module first determines the simulation operation and maintenance recommended strategy of the operation and maintenance demand customer set ranked first, obtains the operation and maintenance root cause residual values of each customer in the operation and maintenance demand customer set, sorts each customer from large to small according to the numerical value of the operation and maintenance root cause residual value, and intervenes in the operation and maintenance management of each customer in sequence according to the sorting. When all customers in the operation and maintenance demand customer set have completed the operation and maintenance management, the simulation operation and maintenance recommended strategy of the operation and maintenance demand customer set ranked next is determined, and the cycle is repeated. The method for obtaining the root cause of the customer's operation and maintenance residual value: select a customer, obtain the simulation operation and maintenance recommendation strategy of the customer's operation and maintenance demand customer set in the last period, respectively extract the simulation operation and maintenance recommendation strategy in the current period and the simulation operation and maintenance recommendation strategy in the last period, extract the simulation operation and maintenance strategy features in the current period and the simulation operation and maintenance strategy features in the last period, match the simulation operation and maintenance strategy features in the current period and the simulation operation and maintenance strategy features in the last period into a simulation operation and maintenance feature group, import the simulation operation and maintenance feature group into the operation and maintenance strategy comparison model, and the operation and maintenance strategy comparison model exports an operation and maintenance strategy residual index AD(s). Set the operation and maintenance demand comprehensive prediction threshold index. When the operation and maintenance demand comprehensive prediction index of the operation and maintenance demand customer set where the customer's operation and maintenance demand prediction period is located is greater than or equal to the operation and maintenance demand comprehensive prediction threshold index, mark the corresponding customer's operation and maintenance demand prediction period as an urgent demand period, and increase the urgent demand number by one. Compare all urgent demand periods two by two. When the two compared urgent demand periods are adjacent customer operation and maintenance demand prediction periods, increase the urgent demand duration number by one. Finally, mark the urgent demand number as KS(c) and the urgent demand duration number as RY(u). Through Calculate the actual operation and maintenance possible demand value of the customer . 2.The AI-based enterprise customer operation and maintenance demand prediction management system according to claim 1, characterized in that, The generation method of a type of operation and maintenance associated prediction fusion feature of an operation and maintenance demand customer set: all customers included in an operation and maintenance demand customer set are determined, the stereoscopic operation and maintenance feature prediction set of each customer is obtained, a type of operation and maintenance associated feature is selected, the type of operation and maintenance associated feature in each stereoscopic operation and maintenance feature prediction set is extracted, all extracted operation and maintenance associated features of the type are fused and processed to generate the type of operation and maintenance associated prediction fusion feature. 3.The AI-based enterprise customer operation and maintenance demand prediction management system according to claim 1, characterized in that, The determination manner of the operation and maintenance demand comprehensive prediction index of the operation and maintenance demand customer set: controlling an operation and maintenance demand unified prediction model of an operation and maintenance demand customer set to carry out T mi length simulation, T mi After the T length simulation is completed, the operation and maintenance demand unified prediction model outputs each type of core physical quantity of the equipment, feature extraction is carried out on each type of core physical quantity, each type of core physical feature is extracted, each type of core physical feature is combined into a comprehensive core physical feature set in a feature set manner, the comprehensive core physical feature set is imported into the operation and maintenance demand model, and the operation and maintenance demand model exports an operation and maintenance demand comprehensive prediction index. 4.The AI-based enterprise customer operation and maintenance demand prediction management system of claim 1, wherein, The generation method of a customer set sequence: each operation and maintenance demand customer set is sorted from large to small according to the numerical value of the operation and maintenance demand comprehensive prediction index, and a customer set sequence is generated according to the sorting. 5.The AI-based enterprise customer operation and maintenance demand prediction management system according to claim 1, characterized in that, The determination method of the simulation operation and maintenance recommended strategy of an operation and maintenance demand customer set: the operation and maintenance demand unified prediction model corresponding to the operation and maintenance demand customer set is simulated multiple times, and the operation and maintenance strategy corresponding to the simulation with the smallest operation and maintenance demand comprehensive prediction index is marked as the simulation operation and maintenance recommended strategy.

6. The AI-based enterprise customer operation and maintenance demand prediction management method is applied to the AI-based enterprise customer operation and maintenance demand prediction management system of any one of claims 1-5, characterized in that, The steps are as follows: Step one: periodically determine the stereoscopic operation and maintenance feature prediction set of each customer of an enterprise, and further generate a plurality of operation and maintenance demand customer sets. Step two: control each operation and maintenance demand customer set to generate various types of operation and maintenance correlation prediction fusion features, and build an operation and maintenance demand unified prediction model for each operation and maintenance demand customer set; Step three: determine the operation and maintenance demand comprehensive prediction index of each operation and maintenance demand customer set, and generate a customer set sequence; Step four: determine the simulation operation and maintenance recommendation strategy of the operation and maintenance demand customer set with the highest ranking, and obtain the operation and maintenance root cause residual value of each customer in the operation and maintenance demand customer set; Step five: sort each customer from large to small according to the numerical value of the operation and maintenance root cause residual value, and intervene in the operation and maintenance management of each customer according to the sorting.

Images