BUSINESS IMPACT

Jun 01, 2021

Time Series Forecasting with Azure Analytic Services

Megan Quinn Posted by Megan Quinn

Despite the ubiquitous impact of time series on a variety of business scenarios, time dependent data can pose a challenge when it comes to modeling and forecasting. Typical regression or classification models are trained on input data constructed of previous observations captured from a single point in time or cross-sectional, meaning time is irrelevant. However, in a time series data set, the order and the intervals at which the data is collected is important. In other words, time is an independent variable, while the value associated with the time is the dependent variable. Examples of this type of data include energy resource prices, stocks, transportation demand, and retail sales as in the graph below displaying furniture sales.

 

In this video, we will use the monthly prices of metals to demonstrate time series forecasting via a Microsoft Accelerator. From accessing data, feature engineering, model validating, and finally - to storing the results, we will walk through all the major steps required to achieve an accurate time series forecasting solution. The highlight, however, is that by leveraging Azure Analytic Services, minimal coding and time series domain knowledge is required for implementation.

 

The resources utilized in this video include Azure Synapse Analytics, Azure Cognitive Services, and Azure Machine Learning. Azure Synapse Analytics is a service that unifies data integration and warehousing with data analytics. It also supports integration with multiple other Azure resources. For our solution, we will use Azure Synapse to not only store data, but access Azure Cognitive Services and Azure Machine Learning, tools that provide the power of AI without expertise knowledge. Through the anomaly detector feature of Azure Cognitive Services, we will evaluate the validity of our time series data and through Azure Machine Learning train a model to accurately forecast metal prices for one year into the future.

 


The overall goals of this video are to demonstrate the applicability of time series forecasting within the viewer’s own business use case, as well as illustrate the code-free ease and integration capabilities of Azure Analytic Services.


 

We can help!

BlueGranite offers a variety of resources to help you learn how you can leverage Modern Data Analytics. Those resources include free training events such as:

Please check out our website to learn more or contact us directly to see how we can help you explore your about modern data analytics options.

 

New call-to-action
Megan Quinn

About The Author

Megan Quinn

Megan has expertise in statistical analysis and machine learning as well as statistical theory. Her recent focus has been centered on predictive maintenance for military fleets with a background in education research as well. She is knowledgeable in a variety of analytical tools including Python, R, SQL, and most recently Spark & Databricks.

Latest Posts

New call-to-action