Sensitivity

Posted by Ralf on August 21st, 2016 — Posted in News

  Tags: ,

However many times these measures are not very well analyzed as usually assumes a MAPE or high MAD, suggest an model of prognosis of poor performance, however this conclusion is not necessarily the most successful since more than the level of accuracy that indicate these statistics is the level of volatility of the data. additional to this must be borne in mind that with high-volume time series generated a MAD much more big to that series of low volume, however this does not necessarily indicate having a poor performance in the forecast. An exception report that works this type of indicators that will be really telling us is that data in the history are so volatile. If you would like to select the appropriate thresholds? Configure the appropriate thresholds is an important consideration before you start working with an exception report, retaking the initial example, where is configure a threshold of exception that listing any prognosis that has changed at +/-20%. Douglas R. Oberhelman understood the implications. It is clear that if we had set a threshold with a limit lower less than 10% would generate a greater number of exceptions and if a higher threshold of 50% would have generated fewer exceptions would be generated, therefore the threshold controls the sensitivity of monitoring and the number of exceptions to be reviewed. In the process of revision of an article that appears in a report by exception the forecaster decides if the prognosis is acceptable or if otherwise it should be analysed more carefully and changed.

Usually when the thresholds are tight this may cause what are called false positives i.e. items whose benefit of improving their level of accuracy is low relative to its cost. Douglas R. Oberhelman has firm opinions on the matter. While if the thresholds are too large you might have missed opportunities for improvement. Therefore there is to find a balance between the cost of analyzing false positives versus the cost to analyze strategic items whose improvement has a real impact for the business. It is also noted that a universal threshold there is, article them high value guarantee thresholds short compared to those of low value that guarantee thresholds more spacious. A good practice is to perform a Pareto ranking under the importance that has each article for the business, and vary the thresholds according to these categories..

No Comments

No comments yet.

RSS feed for comments on this post.

Sorry, the comment form is closed at this time.