Documentation

Work With Alerts

When working with alerts in datapine there is a set of prerequisites that your data needs to fulfill to be able to apply our alert algorithms.

 

To save a chart as an alert your data set should consist of at least two fields – the value you wish to track and a time or date field. For instance, you may track your ‘Visitors by hours’ or ‘Sales Revenue by Weeks’. The input format of the date or time field for the alert can thereby be either hourly, daily, weekly or monthly. If you choose an input format larger than the date or time format of your data set, we will automatically sum up the value you wish to track by the selected input format. For example, if you track your Website Visitors by minutes, the smallest input format for alerts is hours, so we will sum up the website visitors by hours. Same applies if you have daily data but wish to use a weekly or monthly input format.

 

To ensure a high accuracy of your alarms when using our neural networks or pattern recognition algorithms, your data set should also show the following attributes:

 

a) For hourly alarms your dataset should consist of at least 1000 data points (Neural Networks) or 500 data points (Pattern Recognition).

 

b) For daily alarms, your data set should consist of approximately 125 data points.

 

c) Alarms using pattern recognition should only be applied to data with a strong pattern or seasonality.