Research And Implementation Of Field Equipment Data Abnormal Detection Method

In recent years, the rapid development of information technology, data acquisition and storage technology continues to progress, national defense science and technology, accelerating the pace of modernization, and as the proposed of "smart defense" concept, the requirements for military applications of information technology are also increasing. Every country which costs in military field brings the military into booming, at the moment, the field has accumulated a large number of complex data structures. So how to extract valuable knowledge from massive data quickly and efficiently becomes the research focus in the field in recent years. Data mining is proposed effective technique potentially valuable knowledge. With the increased necessity of abnormal data mining, there pays more and more attention to outliers mining tasks. Since the military field has its own characteristics so that this field’s data has an obvious timing characteristics. As a difficult and important research direction of data mining, timing series outlier detection has become a hot spot in recent years.This article study the operation quality evaluation problem of military field control target by the depth analysis of operational data, from the perspective of data mining, identify data anomalies and analyze its causes.First, we get a multidimensional operation timing curve to use timing relationships exist from control target operational data and analysis it. Then, we make sure outlier characterize of control and observation equipment and outlier characterize of operation target, analyze the similarities and differences of them, to build the control target runtime data model through a lot of investigation and analysis of timing relationships graph. Finally, we propose a control target operation quality evaluation algorithm based on the control target runtime data model. This algorithm based on Decision-tree model and evaluation the control target operation quality from coarse-grained level and fine-grained level. At the same time, we designed a series of evaluation to improve the evaluation algorithm, which provides a reference for control target operational assessments.We performed a series of experiments on a real data set provided by the military field to prove that the proposed data model can identify outliers quickly and accurately, which describes the outlier characteristics of observation control equipment better and makes a distinction between outlier of observation control equipment and outlier of control target. In addition, the experiment results also show that it is the same basically between the control target evaluation algorithm of this article proposed and result of the experts evaluate to the control target in reality, and the time of algorithm is smaller than the time of expert evaluation.This article establish operational data model, propose control target operation quality evaluation algorithm, solving commendably the problem of the control target operation quality evaluation rely solely on the expert experience, which makes the auto evaluation possible.