Forecasting (Beta)
This page describes the Automated Forecasting capability.
Thanks to its analytical capabilities, Crystal can explain what happened to your data in the past and in the present (see Automated Insights / Insights Analysis and Alerts respectively).
But what about the future?
With the Automated Forecasting capability, Crystal can also foresee the evolutions in time of your Topics, enabling you to make better decisions by anticipating times!
Pills of Data Analysis
Automated Forecasting is a Data Analysis technique that is part of the Predictive Analytics branch and uses Machine Learning to predict how trends might evolve in the future based on historical data.
This kind of analysis can be applied only to Topics that track a single Entity progress in time, i.e. the Overtime Single Line ones.
Remember
This feature is still in Beta
Request a Forecast
You can request a Forecast by setting it up from a Topic Card.
Set from Topic Card
You can request a Forecast from the Topic Card, like this:
ask for the Topic you want to monitor
once the requested Topic is returned in the Advisor, navigate through the Topic Card and click on the three-dots menu: there, choose the "Forecast" icon
Check for the Activation
At this point, Crystal will give you a feedback about the result of your request.
Once the Forecast has been requested, you simply need to wait until the analysis is available.
Receive a Forecast Notification
When the forecast is available, you'll receive a notification in the Notification Center.
Let's see what happens in this case:
when the notification arrives, open the Notification Centre and read the notification: this notification signals that the forecast is ready
to know more, simply tap on the notification: in this way, you will be redirected to the Advisor
you will now see the Topic in focus with an intro message that tells you that the forecast is ready for your Topic: take a look and see how it will change in the future!
Notice that the results of your forecast include three different lines: "Forecast", "Upper bound", and "Lower bound". The actual forecast is represented by the central line ("Forecast"), while the other lines ("Upper bound" and "Lower bound") serve as confidence intervals: i.e. the elements that define the range within which the actual forecast is confidently expected to lie.
This is it: now you are able to make better decisions anticipating times!
Remember
Once received a Forecast, you can share it with your colleagues from the "Share" button on the Topic Card.
However, as of now, the other actions of the Topic Card are not available for Topics containing a Forecast.
How Does It Work?
Let's briefly see how Crystal can make predictions.
Once a Forecast is requested, Crystal immediately starts to collect the historical data behind that Topic and is able to make statistical calculations around it to estimate how it might evolve.
This is why the final result includes both the actual forecast and the confidence intervals, which define the range within which the actual forecast is confidently expected to lie.
Rules
This feature is available only for Overtime Single Line Topics
In order to guarantee the maximum accuracy and reliability of the resulting predictions, each Topic must meet some conditions of minimal data points presence
The horizon of the future data points predictable is dynamic, algorithmically declared based on the number of data points available
In order to make up for potential data incompleteness, especially in monthly or daily Topics, the statistical model omits the initial and final data points when the metric aggregation function is set to sum or count (including these points in the analysis, particularly when there's only partial data for the last month, could lead to misleading predictions)
If there are Dynamic Filters applied to the Topic, the Forecast generated will take them into account
Limitations
This feature is still in Beta Version
Keep analyzing your data!
Last updated
