XBUS-515 - Time Series Analyses and Databases

Course Description

Understanding how systems and organizations change over time is a foundational part of forecasting and predictive analytics and so univariate and multivariate time series analyses must be a standard part of the data science toolkit. Like most data sets, time series analyses come with their own unique challenges: from handling seasonality and cyclicity to dealing with random walk and structural characteristics to determining methods best suited for dense or sparse data analysis. No two time series data sets are the same and the challenge for the data scientist is to determine how to characterize and treat the data whether the analytical outcome is forecasting, event detection, or visualization and interpretation. 

This course provides a comprehensive view of time series analytics to equip students with the ability to characterize time series data sets and employ state-of-the-art machine learning and analytics tasks for a variety of use cases. We start with a description of various types of time series structures from the regular to the random and describe ways to normalize, smooth, and resample data. The course then focuses on dense time series data (e.g. large amounts of samples at a routine sampling rate) and investigates the use of specialized time series databases for storage and query performance. Finally we will engage several IoT and behavior analytics time series datasets with statistical and machine learning methods for forecasting and anomaly detection, including the use of deep learning mechanisms specifically designed for time series. 

This course is part of the Data Engineering and Data Science tracks of the Advanced Data Science Certificate.

Notes

Enrollment in this course is open to all students and applies credit toward the Advanced Data Science Certificate Data Engineering and Data Science tracks.

Course Objectives

Upon successful completion of the course, students will be able to:

• Describe the design space and requirements of time series databases.

• Compare and contrast InfluxDB, TimescaleDB, and BTrDB for time series data storage.

• Utilize the Python Pandas library for time series data analysis.

• Employ common statistical techniques such as ARIMA, lag analysis, seasonality, and the Holt-Winters method for time series forecasting. 

• Explore the use of RNNs and LSTMs for time series forecasting.

• Discover techniques for event identification and anomaly detection on time series data.

• Utilize ensemble based methods and Scikit-Learn for forecasting and classification.

• Create hierarchical visualizations of time series data and scaling time granularities. 

• Engage visual analytics workflows for time series data including brushing, blending, and aggregation.