Study On Weighted Time Series Prediction Model In Mining Safety System

Prediction technology is the foundation of safety science and technology. Accidentprediction is increasingly becoming the hot issue in the system security management，which isthe main basis for accident prevention and control. Also，accident prediction is the basis offormulating safety production goals and tasks.Statistic predictive model of time series is the main method of accident prediction. Theexisting method of accident prediction did not take full advantage of accident statistical data，that the results can not reflect the actual situation of the security system. The accident timeseries in mining safety system has a typical dynamic nonlinear trend. Weighted prediction wasproposed to characterize time effect of accident time series in mining safety system，in whichthe larger weights was given to recently statistical data while smaller weights to longer-term.Based on the exponential function，weighted time series prediction model was established foraccident in safety system. As needed，the short-term testing of accuracy was formed forweighted time series prediction model. And then， this paper intend to generalize thecharacteristic attribute of five weighting functions，including time nonlinear function，modifiedexponential function，statistics-nonlinear function，residuals-time nonlinear function，and Pierrefunction. Based on a modified exponential function，the discriminate function and selectionmethod of weighting function were constructed. According to the statistical data of accident inmining safety system in china from2005to2009，the accuracy of weighted prediction is higherthan the traditional exponential prediction model. Specially，the result of weighted prediction onresiduals-time nonlinear function had the highest accuracy that was70.8094. Compared withthe exponential model，the accuracy of weighting predication was improved of824.26%.Further more，the influence with the average growth rate of modified exponential function usingthe discriminate function was that the prediction accuracy increases with the average growthrate in weighted prediction. At the same time，the accuracy of the Pierre function as weightingpredication is65.4270，as well as improved of900.29%. The recently accident predictionaccuracy was improved in weighted prediction. It’s proved that the weighted predication modeltally with the actual situation well，which provides new ideas for the accident predictiontheories and methods. There is an important practical significance for safety management.