So far, we've used TensorFlow mostly for image processing, and, to a lesser extent, for text sequence processing. In this chapter, we will tackle a specific type of tabular data: time-series, data.

The time series data comes from many domains with usually one commonality—the only field changing constantly is a time or sequence field. It is common in a variety of fields, but especially common in economics, finance, health, medicine, environmental engineering, and control engineering. We'll dive into examples throughout the chapter, but the key thing to remember is that order matters. Unlike in previous chapters, where we shuffled our data freely, time series data cannot be shuffled that way without losing meaning. An added complexity can be the availability of data itself; if we have data available up until the current time with no further historical data to capture, no amount of data collection will possibly produce more—you are bound by time-based availability.

Luckily, we're going to dive into an area with copious amounts of data: the financial world. We'll explore some types of things hedge funds and other sophisticated investors may do with a time series data.

In this chapter, we will cover the following topics:

• What a time series data is and its special properties
• Types of input and approaches investments firms may use in their quantitative and ML-driven investment efforts
• Financial time series data and how it is obtained; we'll obtain some live financial data as well
• The application of modified convolutional neural networks to finance