From the course: Applied AI for IT Operations (AIOps)

Time series forecasting - Python Tutorial

From the course: Applied AI for IT Operations (AIOps)

Start my 1-month free trial

Time series forecasting

“

- [Instructor] In this chapter, I will demonstrate how to forecast time series data for IT Ops use cases. What is a time series? A time series is a series of data points indexed in time order. It is a sequence of samples taken at successfully, equally spaced points in time. For example, this may include daily closing stock prices, hourly peak concurrent users in a website and total emails every minute in an email server. Time series data may be a steady sequence or may exhibit multiple trends. These trends can be cyclic, growth or seasonal. Examples of cyclic trends include hourly concurrent users in a website. Typically, the peak is much higher during midday than during night. Seasonal trends include higher daily travelers during holiday seasons. Trying to predict time series data for the future is a unique problem, because the model should account for the cyclic, growth and seasonal trends that are usually seen in time series data. What is service load forecasting? A service in IT ops is any service that provides a specific function. Examples include websites, databases, APIs, et cetera. Each of these services need to be provided resources to handle the expected volumes of data. Providing too many resources means waste of money, providing too few resources would result in service degradation and failure. Service load forecasting predicts expected future workload by time. It plays a key role in creating the right capacity to handle these loads. The load pattern on any service is usually a time series, and the load varies based on the day of the week and time of the day. Typical corporate services have higher loads during midday and lower loads during nights and weekends. Time series forecasting involves predicting future values by time, based on previously observed values, There are multiple use cases for time series forecasting in IT ops. This includes website request loads, storage usage for databases, transaction loads on business services, et cetera. In this chapter, we will predict the future time series values for website request loads. Before that, in the next video, let's review recurrent neural networks.

Contents