Research Of Mining Algorithms Of Intelligence Data Based On Support Vector Machine

With the rapid development of industrialization, machinery and equipment aremoving towards large-scale, high-efficiency, high intelligent direction. Linkages betweendifferent components of mechanical equipment increasingly close, and any of acomponent malfunction or mutation can cause major accidents, resulting from theproduction process interrupting. How to ensure machinery and equipment, safety, stablyand efficiently operation is becoming today’s scholars’ major issue to study. Theassessment of the state of faulty equipment effective, fast, and accurately can findequipment malfunction fast, and timely and effectively deal with, which is important tothe prevention of major accidents.Support vector machines have not only better results when treated with learningproblems, high latitude, nonlinear, and small samples, but also has a strong ability topromote. Currently, support vector machines are widely used to identify the fault, thedata forecasting and data mining. The main work is as follows:Proposed fault identification method based on EEMD-SVM’s. As a feature extractionmethod, EMD exists aliasing phenomenon in processing signals and largely affects thevalidity and authenticity of the characteristics. Therefore, in order to better extract theSVM features, The improved EMD method is selected that is EEMD. It is a good way toanti-alias phenomenon EMD appear. And through experiments prove the accuracy of themethod.The SVM is applied to predict aspects of data. It includes periodic signals andprediction forecast for non-periodic signals. For these two different characteristics of thesignal, the paper adopts a different approach in conjunction with SVM and verifies byexamples illustrate the effectiveness of the two methods of SVM. In the periodic signal，SVM combines the phase space reconstruction algorithm optimized by PSO; In thenon-periodic signal prediction process, in order to better achieve predictable results, theSVM with combine PSO and EEMD. The examples demonstrate that, whether periodic ornon-periodic signal prediction, SVM can achieve good results.SVM can deal with the small sample of events, so this paper proposes mining datamodel of pump head based on support vector machine. Through the limited measurement point training, the use of SVM can dig out as much of the mid-point measurement pointsto accurately draw trend curve.