1 00:00:11,120 --> 00:00:15,410 So in this lecture, we are going to summarize everything we learned in this section. 2 00:00:16,280 --> 00:00:21,080 This section looked at times there is basics which was designed to prepare you for the rest of the course. 3 00:00:21,590 --> 00:00:23,900 We first started with a very simple question. 4 00:00:24,590 --> 00:00:25,970 What is a time series? 5 00:00:26,400 --> 00:00:31,280 Hopefully you now have a better understanding of what a time series is and what constraints must be 6 00:00:31,280 --> 00:00:34,120 met so that they apply to the methods we will study. 7 00:00:35,060 --> 00:00:38,130 We next looked at the concept of modeling versus prediction. 8 00:00:38,660 --> 00:00:44,330 This is an important distinction because sometimes we'll be focused on one and other times will be focused 9 00:00:44,330 --> 00:00:45,010 on the other. 10 00:00:45,620 --> 00:00:51,740 You should now understand why modeling is useful and in fact can aid and prediction, especially when 11 00:00:51,740 --> 00:00:54,740 it gives you hints about when prediction is a waste of time. 12 00:00:55,880 --> 00:01:01,250 The next step was to consider the shape of our data sets in this course, since we'll be working with 13 00:01:01,250 --> 00:01:02,960 Python and Python libraries. 14 00:01:03,230 --> 00:01:05,780 There are certain conventions that we normally adhere to. 15 00:01:06,350 --> 00:01:11,990 You should also now understand why having this reflex of automatically thinking about shapes will help 16 00:01:11,990 --> 00:01:13,720 your intuition in machine learning. 17 00:01:15,350 --> 00:01:18,440 The next step was to discuss different time series tasks. 18 00:01:18,920 --> 00:01:25,080 We looked at the one step forecast, the multi-step forecast and the multi output multi-step forecast. 19 00:01:25,640 --> 00:01:31,250 We also considered the practical applications of time series classification, which we will study in 20 00:01:31,250 --> 00:01:32,030 this course. 21 00:01:33,280 --> 00:01:39,040 The next step was to learn about common data transformations used in Time series, the transformations 22 00:01:39,040 --> 00:01:43,750 we studied all have a common property, which is that they tend to squash large values. 23 00:01:44,140 --> 00:01:49,090 We learned about why this is a deeply fundamental operation reflected in nature itself. 24 00:01:49,990 --> 00:01:55,030 We next learned about Time series metrics, which are more numerous compared to other fields of data 25 00:01:55,030 --> 00:01:56,950 analysis and machine learning. 26 00:01:57,760 --> 00:02:03,370 This also taught us about one of the major themes of this course, which is that there are so many options 27 00:02:03,370 --> 00:02:04,120 to choose from. 28 00:02:05,140 --> 00:02:09,730 Exercises for this course are essentially built in because there are many combinations. 29 00:02:09,730 --> 00:02:11,740 We simply do not have time to try. 30 00:02:12,280 --> 00:02:17,440 Your job in this course will be to try different combinations you think might work and see if you can 31 00:02:17,440 --> 00:02:18,930 obtain a superior result. 32 00:02:23,660 --> 00:02:29,030 The next step for us was to turn our attention to Financial Times series, we learned about important 33 00:02:29,030 --> 00:02:32,240 concepts such as the log price and the log return. 34 00:02:33,170 --> 00:02:38,510 We then had a chance to do some practical work which involved simulating a stock price time series. 35 00:02:39,380 --> 00:02:44,240 In the next lecture, we learned about why that practical work was useful when we studied the random 36 00:02:44,240 --> 00:02:46,170 walk hypothesis in depth. 37 00:02:46,730 --> 00:02:52,070 We then learned about the naive forecasts and why it's extremely important to have baseline predictions 38 00:02:53,000 --> 00:02:54,600 for random walks in particular. 39 00:02:54,740 --> 00:02:58,370 It turns out that the naive forecast is the best forecast. 40 00:02:59,330 --> 00:03:05,390 The final exercise for this section was to implement the new forecasts and to test out different metrics. 41 00:03:06,470 --> 00:03:09,620 So hopefully you learned a lot and we'll see you in the next lecture.