Closed loop management system algorithms of acquisition operations and maintenance based on alternation early warning

Abstract

In order to construct the closed-loop management system of abnormal early warning, the decision tree algorithm, survival analysis algorithm and logistic regression algorithm are used synthetically. The improved logistic regression algorithm proposed in this study is used to establish the abnormal early warning model, identify the abnormal tendency in advance, and construct the active monitoring and closed-loop operation and maintenance management system to provide the key technical support. The results show that this method improves the real-time processing of data acquisition, reduces the impact on data acquisition, improves the efficiency of operation and maintenance, and carries out timely and accurate early warning. This method proposes active monitoring mechanism based on complex event processing and fast fault location technology based on expert database. From this point of view, testing the key technical indicators of the system proves that the system can effectively promote the improvement of business operation and maintenance efficiency, quality and management level.

Keywords

Alteration early warning closed-loop management system operation and maintenance management information collection

1. Introduction

With the rapid increase of data volume, in order to improve system throughput, the efficiency optimization of huge data will require higher performance of data processing. In the process of querying or analyzing complex data, if one data fails, the whole query process does not need to be repeated. With the increase of the number of power information acquisition terminals, it will inevitably increase the probability of failure [1]. In the case of large-scale data acquisition, data failure is no longer a rare event. (According to Google’s test report, in the MapReduce data processing tasks, each task will have 1.1–1.2 data node failure.) Therefore, in large-scale complex data environment, it is unrealistic to rely solely on hardware, and the system should consider software-level fault tolerance more. Due to the asynchronism and difference of the construction of the acquisition system, the support of heterogeneous environment for data generated under different acquisition devices and hardware environments can effectively reduce the requirement of hardware input, reduce the construction cost, and play a good role in load balancing and task scheduling of different node performance [2, 3, 4]. According to the operation and maintenance characteristics of information acquisition, data analysis is required to reduce data preparation time as much as possible, and the system can respond quickly to the data requirements and conduct data analysis.

On the premise of meeting the system requirements, in view of China’s national conditions, it is an important indicator to build a data analysis system at a lower cost. It is not only the input of hardware or software, but also the daily operation and maintenance of the system. Cost reduction can make the system realize in practical application as soon as possible.

2. Methodology

Constructing the closed-loop management system of customer dislocation is mainly guided by the theory of “critical moment of customer life cycle recession”. It synthetically uses “survival analysis data mining algorithm”, “logical regression data mining algorithm” and “decision tree data mining algorithm”, establishes the customer dislocation early warning model, and finds out the customer dislocation tendency in advance. By using “two-step clustering data mining algorithm” combined with customer behavior attributes and research results, a customer segmentation model is established, which divides customers into five groups, and realizes the location of customer dislocation reasons [5, 6]. Combined with customer abnormal early warning model and customer segmentation model, five kinds of customer abnormal early warning model are constructed. Finally, a personalized customer abnormal early warning model is constructed by using the five kinds of early warning model synthetically, so as to establish a more accurate customer abnormal early warning model, and locate the cause of abnormal changes according to the pertinent early warning information provided by the early warning model, so as to realize the accurate early warning of customers.

Customer life cycle theory, also known as customer relationship life cycle theory, refers to a customer’s birth, growth, maturity, aging and death process similar to life for an enterprise. It is the development track of customer relationship level which changes with time. It dynamically describes the general characteristics of customer relationship in different stages [7]. Specifically, to different industries, there are different detailed definitions of this. For example, in the telecommunications industry, the so-called customer life cycle refers to the process of telecom customers from becoming customers of telecommunications companies and starting to produce business consumption, consumption growth, consumption stability, consumption decline, and finally leaving the network, as shown in Table 1.

Table 1
Theory table of customer life cycle

	Stage A: user acquisition	Stage B: user build-up	Stage C: user climax	Stage D: user decline	Stage E: user exit
Issues to be considered in this life cycle phase	$\cdot$ How to find and obtain potential users?	$\cdot$ How to cultivate users into high-value users?	$\cdot$ How can users use new products? $\cdot$ How to cultivate customer loyalty?	$\cdot$ How to prolong the user’s “life cycle”?	$\cdot$ How to win back users?
Key work points of the period	$\cdot$ Attracting customer network	$\cdot$ Product portfolio/ services to stimulate demand	$\cdot$ Cross selling $\cdot$ Targeted marketing $\cdot$ Differentiated services for high value users	$\cdot$ High risk user oath mechanism $\cdot$ High-risk user retention measures	$\cdot$ Win back the high value method

When the customer enters the off-line period and then judges the customer’s dislocation tendency, although the accuracy is high, the success rate of retaining the dislocation user is low. It is equivalent to the medical diagnosis of cancer patients, once diagnosed as advanced, the success rate of cure is relatively low [8]. Therefore, guided by the theory of “critical time of customer life cycle recession”, this study warns customers in the recession period before customer dislocation, discovers user dislocation tendency in advance, and predicts user dislocation time. Decision tree prediction model: Decision tree is a case-based inductive learning algorithm, which focuses on inferring classification rules represented by decision tree from a set of disordered and irregular examples [9]. It uses a top-down recursive method to compare the attributes of the internal nodes of the decision tree, and judges the branches from the node down according to the different attribute values, and gets the conclusion at the leaf nodes of the decision tree, as shown in Fig. 1.

Figure 1.

Schematic diagram of node composition of decision tree method.

Decision tree algorithm is a method to approximate the value of discrete function. It is a typical classification method. Firstly, the data are processed, and the readable rules and decision trees are generated by induction algorithm. Then, the new data are analyzed by decision. In essence, decision tree is a process of classifying data through a series of rules. It can be accomplished by training set $Y$ . $Y=\left\{{<x_{i},c_{i}>}\right\}$ . $x_{i}=\left({a_{1},a_{2},\cdots,a_{n}}\right)$ is a training instance. They have n attributes listed in attribute table $\left({A_{1},A_{2},\cdots,A_{n}}\right)$ , where $a_{1}$ represents the value of attribute $A_{i}$ . $c_{i}\in C=\left\{{C_{1},C_{2},\cdots,C_{m}}\right\}$ is the classification result of $X$ .

Logical regression prediction model, is similar to linear regression, but the target field uses character field instead of numerical field. Logistic regression establishes a set of equations that relate the input attribute values to the probabilities of each class of output fields. Once the model is generated, it can be used to estimate the probability that a new record belongs to a certain class. The target class with the highest probability is designated as the predicted output value of the record. The main application of logistic regression in the industry is classified prediction. The most commonly used case is to predict the probability of alienated off-line users leaving the network. Logistic regression dependent variables can be bi-categorized or multi-categorized, but bi-categorization is more common and easier to explain. Logistic regression algorithm is introduced with two-class Logistic regression as an example. The dependent variable $Y$ is set as a binary variable, and its values are $Y=$ 1 and $Y=$ 0. The m independent variables affecting the value of $Y$ are $X_{1},X_{2},\cdots,X_{m}$ . When the conditional probability of event occurrence under the action of M independent variables is $P=P\left({Y=1|{X_{1},X_{2},\cdots,X_{m}}}\right)$ , the logistic regression model can be expressed as follows:

$\displaystyle P=\frac{\exp\left({\beta_{0}+\beta_{1}X_{1}+\beta_{2}X_{2}+% \cdots+\beta_{m}X_{m}}\right)}{1+\exp\left({\beta_{0}+\beta_{1}X_{1}+\beta_{2}% X_{2}+\cdots+\beta_{m}X_{m}}\right)}$ (1)

Among them, $\beta_{0}$ is a constant term and $\beta_{1},\beta_{2},\cdots,\beta_{m}$ is a partial regression coefficient.

$\displaystyle\ln\frac{P}{1-P}=\ln\frac{\frac{\exp\left({\beta_{0}+\beta_{1}X_{% 1}+\beta_{2}X_{2}+\cdots+\beta_{m}X_{m}}\right)}{1+\exp\left({\beta_{0}+\beta_% {1}X_{1}+\beta_{2}X_{2}+\cdots+\beta_{m}X_{m}}\right)}}{1-\frac{\exp\left({% \beta_{0}+\beta_{1}X_{1}+\beta_{2}X_{2}+\cdots+\beta_{m}X_{m}}\right)}{1+\exp% \left({\beta_{0}+\beta_{1}X_{1}+\beta_{2}X_{2}+\cdots+\beta_{m}X_{m}}\right)}}$ (2)

Logic regression algorithm has the advantage of minimizing the difference between the accuracy of training set and test set. It can meet the requirements of logistic regression algorithm by missing value processing and correlation analysis.

3. Results and discussion

3.1 System function design

The closed-loop management system of customer dislocation consists of three subsystems. Based on the analysis results of system function of system demand analysis, the overall functional structure of the system is designed, as shown in Fig. 2.

Figure 2.

The overall functional structure of the system.

The subsystem of early warning worksheet management has three main system functions: scoring function of abnormal early warning model, setting function of abnormal early warning rules, and inquiry function of abnormal early warning. It provides three man-machine interfaces, as shown in Fig. 3.

Figure 3.

Overall functional structure diagram of abnormal motion early warning subsystem.

The scoring function of abnormal early warning model of abnormal early warning subsystem is that the background automatic operation function does not provide man-machine interface. The setting function of abnormal early warning rules provides one man-machine interface, and the abnormal early warning query function provides detailed list query and summary query with two man-machine interfaces. The function flow chart of the alternation early warning subsystem is shown in Fig. 4.

Figure 4.

Functional flow chart of alien early warning subsystem.

3.2 Active monitoring and closed loop operation and maintenance management

Business early warning and online diagnosis subsystem realizes business early warning of system abnormalities and online diagnosis and analysis of system faults. It includes monitoring module, expert system reasoning engine and expert knowledge base. The monitoring module effectively monitors various equipment and applications by establishing a full-dimensional data monitoring model, discovers data abnormalities in time, and starts business early warning and online diagnosis. Expert system inference machine realizes on-line diagnosis and analysis of abnormal business warning and system failure by using fault diagnosis rules and business early warning rules in knowledge base. Expert knowledge stock is the knowledge of rules needed in business early warning and online diagnosis which is summarized by theoretical research and practical operation and maintenance experience of experts. These rules can be amended according to the quality of operation and maintenance.

The functions of the closed-loop operation and maintenance management subsystem include closed-loop management, comprehensive evaluation and operation and maintenance learning management. Closed-loop operation and maintenance management realizes the whole closed-loop monitoring and management of discovery, diagnosis, disposal, review and traceability of the complete “operation and maintenance life cycle”, and can timely monitor, assess and supervise the quality and efficiency of operation and maintenance work. It realizes the operability, scientificity and standardization of performance evaluation of operation and maintenance work, and applies multi-dimensional analysis technology to analyze and classify operation and maintenance data from multiple dimensions, realizes visualization and visualization of operation and maintenance work results display. Comprehensive evaluation realizes the operability, scientificity and standardization of performance evaluation of operation and maintenance work. By using multi-dimensional analysis technology, the operation and maintenance data are analyzed and classified from various dimensions, and the visualization and visualization of the operation and maintenance work results are realized. Operations and maintenance learning management realizes the effective precipitation and accumulation of operation and maintenance knowledge and methods acquired in the process of operation and maintenance by screening operation and maintenance cases.

3.3 Early warning test

Intelligent monitoring and management function is introduced into workflow technology, which can monitor and manage the whole process of operation and maintenance of the system through work order management, and realize the fine requirements of operation and maintenance management. The early warning information and fault information are displayed according to the comparison between the index data of the database, server, middleware, acquisition system, pre-communication platform and the index threshold of system maintenance. The main contents are object number, object name, index name, early warning type, early warning threshold, fault threshold, index data, early warning reason and so on. It also provides maintenance and preservation of the causes of early warning. The types of early warning are divided into early warning and failure, and the detailed information of monitoring objects can be viewed.

The “active monitoring and closed-loop operation and maintenance management” is simulated and tested from two aspects of monitoring and early warning. The test proves that it can realize on-line monitoring and early warning of faults, achieve timely and accurate fault early warning, standardize early warning process, and effectively promote the improvement of operational and maintenance efficiency, quality and management level.

4. Conclusion

The construction of fault early warning, timely and accurate, operation and maintenance process specification, and “active monitoring and operation and maintenance closed-loop management” oriented to business indicators are studied. It consists of three sub-frameworks, including information perception framework, business early warning and online diagnosis framework and closed-loop operation and maintenance management framework. In view of the large data analysis and processing in the construction, the exploration of operation and maintenance mode has been deeply studied. The design can effectively promote the improvement of operation and maintenance efficiency, quality and management level.

Operation and maintenance quality is the guarantee of stable and efficient operation of information collection. In order to strengthen the management of operation and maintenance work, continuously improve the level of operation and maintenance, and further improve the quality of operation and maintenance, it is necessary to focus on how to objectively, comprehensively and comprehensively evaluate the quality of operation and maintenance qualitatively and quantitatively. In order to strengthen the management of operation and maintenance work, continuously improve the level of operation and maintenance, and further improve the quality of operation and maintenance, how to objectively, comprehensively and comprehensively evaluate the quality of operation and maintenance needs to be emphatically studied. According to the feedback result of maintenance order and the change result of customer status data, the system evaluates and scores the maintenance effect of the customer manager, pushes the evaluation score to the remuneration management system, and manages the performance salary of the customer manager in a unified way. It solidifies the idea of maintaining the management process of customer’ s abnormal closed-loop into the system, and completes the design and implementation of the customer’ s abnormal closed-loop management system.

References

Hayes

B.P.

and Gruber

J.K.

, Prodanovic M. A closed-loop state estimation tool for MV network monitoring and operation, IEEE Transactions on Smart Grid6(4) (2017), 2116–2125.

Qiu

Zhang

and Wei

, Feasibility study of closed-loop no-insulation coil to shield alternating magnetic field, IEEE Transactions on Applied Superconductivity27(6) (2017), 1–5.

Liu

Zhang

and Guo

, Damping characteristics of magneto-rheological fluid absorber based on closed-loop proportion integration differentiation control, Materialwissenschaft Und Werkstofftechnik49(2) (2018), 235–239.

Mascio

F.D.

Galli

and Natalini

, Learning-shaping crises: A longitudinal comparison of public personnel reforms in italy, 1992–2014, Journal of Comparative Policy Analysis19(2) (2017), 1–20.

Ling

and Proakis

, Synchronization in digital communication systems: basics of phase-locked loop techniques, Journal of Cystic Fibrosis6(7) (2017), S32.

Kulkarni

Lee

S.W.

and Kluck

, Real-time closed loop diastolic interval control prevents cardiac alternans in isolated whole rabbit hearts, Annals of Biomedical Engineering46(11) (2018), 1–12.

Pereira

M.M.

Machado

R.L.

and Pires

S.R.I.

, Forecasting scrap tires returns in closed-loop supply chains in brazil, Journal of Cleaner Production188 (2018), 741–750.

Min

and Shi

, A modal interval based method for dynamic decision model considering uncertain quality of used products in remanufacturing, Journal of Intelligent Manufacturing29(4) (2018), 925–935.

Zhang

and Liu

, Multifunctional cyber-physical system testbed based on a source-grid combined scheduling control simulation system, Iet Generation Transmission & Distribution11(12) (2017), 3144–3151.

Closed loop management system algorithms of acquisition operations and maintenance based on alternation early warning

Abstract

Keywords

1. Introduction

2. Methodology

Table 1 Theory table of customer life cycle

3.1 System function design

3.3 Early warning test

4. Conclusion

References

Table 1
Theory table of customer life cycle