Deploying predictive analytics at the right time
You want to use the predictive analytics capabilities in Splunk ITSI to forecast trends and get ahead of the competition. You want to know when is the right time to deploy this capability so that you use it correctly.
This article is part of the Definitive Guide to Best Practices for IT Service Intelligence. ITSI end users will benefit from adopting this practice as they work on Service Insights.
Solution
To be sure you are ready to begin using predictive analytics, make sure you meet these five key requirements:
- Accurate health service scores. If you don't have accurate historical data to base predictions on, your predictions will be worthless.
- Sufficient training data. At a minimum, you need 14 days of data, but 30, 60, or 90 will be much better.
- Cyclical patterns in KPIs. The training data should contain expected cyclical changes so those don't get flagged as unexpected events.
- Failures in training data. Just as predictive analytics relies on accurate historical data, it also relies on failure data. It can only predict failures that are similar to those seen in the past. Therefore, if your training data is perfectly clean, predicting failures will be difficult because it has nothing to learn from.
- Knowledge of machine learning math. Predicting the future isn't a trivial task. Setting up a good predictive analytics model requires time and knowledge investment. When creating your model, you will have to make decisions you can only make with some understanding of machine learning algorithms.
Next steps
You might also be interested in the following Splunk resources:
- Splunk Docs: Service insights manual
- Splunk Docs: Using the predictive analytics dashboard in ITSI
- Splunk: What is predictive analytics?
- Splunk Blog: Making smarter predictions in ITSI
- App: Smart ITSI Insights App for Splunk