|
| Exploration of Repairing Temperature Monitoring Patch Missing Data with GM （1， 1）Sinusoidal Model |
| Received:September 21, 2014 Revised:February 15, 2015 |
| View Full Text View/Add Comment Download reader |
| DOI：10.7643/issn.1672-9242.2015.01.006 |
| KeyWord:GM（1,1） sinusoidal model timing corrected model standard model model fitting repairing prediction |
| Author | Institution |
| GUO Zan-hong |
South West Institute of Technical Engineering，Chongqing ，China |
| TANG Qi-huan |
South West Institute of Technical Engineering，Chongqing ，China |
|
| Hits: |
| Download times: |
| Abstract: |
| Objective To e :stablishing a method and model with better prediction accuracy for repairing atmospheric temperature monitoring missing data. Methods GM （1，1） standard model was modified by sine function，and the missing data was repaired by the segmented GM（1， 1）sinusoidal model established. Using the temperature diary time value data monitored one day in Wanning station as the test data，the GM（1， 1） standard model，GM（1， 1）timing corrected model and GM（1， 1）sinusoidal model were established at the same time. Missing data repairing by the three models were comparatively analyzed，and the relatively better repair model was determined. Results From the analysis of the model fitting results，the fitting of the GM（1， 1）standard model and GM（1， 1） sinusoidal model was the best，while fitting of the GM（1，1）timing corrected model was relatively poor. From the prediction accuracy of the analysis，GM（1，1）standard model and GM（1，1）timing corrected model showed poor results in repairing prediction，with the average relative errors of 22.54% and 17.70% ，respectively. Whereas the average prediction error of GM（1， 1）sinusoidal model was only 3.14%，which showed great improvement，and had a good prediction result. Conclusion GM（1， 1）sinusoidal model could well repair the missing data，and its result was better than those of GM（1， 1）standard model and GM（1， 1）timing corrected model . |
| Close |