1 00:00:10,950 --> 00:00:13,110 Hey, everyone, and welcome back to the course. 2 00:00:13,110 --> 00:00:16,380 And we are now going to introduce the next section. 3 00:00:16,560 --> 00:00:21,420 So this section is all about time series basics and a primer on finance. 4 00:00:21,420 --> 00:00:27,210 Since I know many of the students in this course are interested in time series analysis specifically 5 00:00:27,210 --> 00:00:28,950 for financial applications. 6 00:00:33,280 --> 00:00:37,390 So let's briefly outline what we will discuss in this section and why. 7 00:00:38,260 --> 00:00:42,040 So we'll start this section by answering a very basic question. 8 00:00:42,220 --> 00:00:44,020 What is a Time series? 9 00:00:44,410 --> 00:00:50,170 Now, although you may have some preexisting intuition about what a Time series is, this will give 10 00:00:50,170 --> 00:00:52,160 us a concrete definition to work with. 11 00:00:52,180 --> 00:00:58,240 In this class, we'll also provide many examples to get you thinking about applications that you may 12 00:00:58,240 --> 00:00:59,710 not have considered before. 13 00:01:00,980 --> 00:01:04,670 Next, we'll discuss the concepts of modeling and predicting. 14 00:01:05,000 --> 00:01:10,760 Now, most new students to machine learning think of these as the same thing, but in fact, they are 15 00:01:10,760 --> 00:01:11,600 separate. 16 00:01:11,840 --> 00:01:14,750 There are times where we care more about one over the other. 17 00:01:15,110 --> 00:01:20,780 It's important to be able to identify each of these tasks and to know when you are doing what. 18 00:01:21,740 --> 00:01:26,780 So the next step is to answer another basic question Why do we care about shapes? 19 00:01:27,110 --> 00:01:32,420 This is a concept that is not covered in most courses in books, but that I feel is very important. 20 00:01:32,600 --> 00:01:37,910 So you'll see this concept a lot in my courses, especially courses about images and sequences. 21 00:01:38,180 --> 00:01:43,520 When you're able to visualize your data automatically because you inherently understand its shape, 22 00:01:43,520 --> 00:01:45,410 this is a powerful reflex. 23 00:01:46,970 --> 00:01:50,450 The next step is to discuss the kinds of time series tasks. 24 00:01:50,750 --> 00:01:56,720 Now, most people think of time series analysis is just one thing, but in fact there are multiple approaches. 25 00:01:56,720 --> 00:02:01,460 We have one set of forecasts, multi step forecasts and multi output forecasts. 26 00:02:01,580 --> 00:02:07,010 We also have time series classification, which is typically not discussed in a traditional time series 27 00:02:07,010 --> 00:02:07,790 course. 28 00:02:08,980 --> 00:02:14,200 The next step is to discuss data transformations specifically of the box Cox variety. 29 00:02:14,620 --> 00:02:20,050 Time series transformations are a bit different from regular machine learning, so you'll want to understand 30 00:02:20,050 --> 00:02:22,630 the common transformations and why we use them. 31 00:02:23,800 --> 00:02:26,920 The next step is to discuss forecasting metrics. 32 00:02:27,400 --> 00:02:31,900 Now, for some reason, Time series analysis has a lot of different metrics. 33 00:02:32,050 --> 00:02:37,930 My personal preference is to just stick to one, but in order to understand papers and communicate with 34 00:02:37,930 --> 00:02:39,970 your colleagues, these are important to know. 35 00:02:40,810 --> 00:02:43,840 After this, we'll do a Financial Time series Primer. 36 00:02:44,050 --> 00:02:50,050 This will introduce you to important quantities specific to finance, such as the net return gross return 37 00:02:50,050 --> 00:02:54,730 log return log prices, adjusted, close dividends and so forth. 38 00:02:54,820 --> 00:03:00,370 So if you want to do finance, although Financial Time series do look like regular time series, you'll 39 00:03:00,370 --> 00:03:01,840 still want to know what these things mean. 40 00:03:03,280 --> 00:03:07,840 Next, we'll have a chance to write a bit of code when we do stock price simulations. 41 00:03:08,170 --> 00:03:13,270 Now, this might seem a little random, but the idea is to get you started early with doing practical 42 00:03:13,270 --> 00:03:16,330 work so that we can build on these skills later in the course. 43 00:03:16,630 --> 00:03:20,590 Speaking of which, the next lecture will build on these price simulations. 44 00:03:21,010 --> 00:03:25,600 At this point, we'll bring things back to the theory world and gain more insight about the meaning 45 00:03:25,600 --> 00:03:27,520 behind those price simulations. 46 00:03:27,850 --> 00:03:31,360 In fact, those price simulations were not that far from reality. 47 00:03:31,630 --> 00:03:35,530 We'll discuss the random walk and the corresponding random walk hypothesis. 48 00:03:35,890 --> 00:03:41,050 This is a very important concept in time series, analysis and finance because it tells you what is 49 00:03:41,050 --> 00:03:44,050 possible and what is impossible to forecast. 50 00:03:44,710 --> 00:03:49,720 In the next lecture, we'll discuss the naive forecast, as well as the importance of having a baseline 51 00:03:49,720 --> 00:03:50,470 prediction. 52 00:03:51,190 --> 00:03:57,370 In practice, the naive forecast often serves as a baseline model with which you can compare other models. 53 00:03:57,700 --> 00:04:02,530 It might seem like a weak tool when you compare it alongside powerful machine learning models, but 54 00:04:02,530 --> 00:04:04,690 you'll be surprised that the results we observe. 55 00:04:06,390 --> 00:04:11,400 The next step in this section will be to do a little more practical work where we learn about how to 56 00:04:11,400 --> 00:04:14,700 implement and evaluate the naive forecast and code. 57 00:04:15,030 --> 00:04:19,770 This will teach us how to actually perform the naive forecast and make use of the metrics we learned 58 00:04:19,770 --> 00:04:20,670 about earlier. 59 00:04:21,450 --> 00:04:23,130 So that's the outline for this section. 60 00:04:23,160 --> 00:04:25,440 Thanks for listening and I'll see you in the next lecture.