1 00:00:00,510 --> 00:00:01,670 Y'all ready for this. 2 00:00:01,680 --> 00:00:05,630 We've got step six which is experimentation. 3 00:00:05,730 --> 00:00:12,390 That's important to remember all of these steps here are a highly iterative process meaning you might 4 00:00:12,390 --> 00:00:19,380 try one thing find out it doesn't work then try another another and another and just keep going through 5 00:00:19,380 --> 00:00:20,160 this loop right. 6 00:00:21,000 --> 00:00:23,520 But how would all of these look in practice. 7 00:00:23,520 --> 00:00:26,940 Well with a process like the one we've got here. 8 00:00:26,940 --> 00:00:32,850 Machine learning projects become a tool matching projects and that's what we're going to do. 9 00:00:32,910 --> 00:00:38,440 We're going to use this process and learn the different tools for each step. 10 00:00:39,400 --> 00:00:45,280 So eventually you'll be able to start to look at a machine learning problem and think about OK I know 11 00:00:45,280 --> 00:00:47,070 what tools I can use for each of these STEM. 12 00:00:47,080 --> 00:00:48,060 We can break it down. 13 00:00:48,070 --> 00:00:55,340 I can use this for that for that and apply this process to whatever kind of problem you're working on. 14 00:00:55,380 --> 00:01:01,850 For example someone might come to you with a dataset and ask you to find insights on it. 15 00:01:01,950 --> 00:01:06,330 You might start by defining a problem within which is step one. 16 00:01:06,510 --> 00:01:10,080 Then you go through to step two and look at the data. 17 00:01:10,080 --> 00:01:12,390 This part is called data analysis. 18 00:01:12,390 --> 00:01:19,230 Here you might define an evaluation metric based on the problem at the same time as inspecting the different 19 00:01:19,230 --> 00:01:23,370 features of the data which is step three step four. 20 00:01:23,610 --> 00:01:29,040 Once you know a little bit more about the data you decide to build a machine learning model using the 21 00:01:29,040 --> 00:01:31,910 features you found to predict some target. 22 00:01:31,910 --> 00:01:34,090 Now this is Step 5. 23 00:01:34,300 --> 00:01:36,080 Your first model goes pretty well. 24 00:01:36,200 --> 00:01:39,960 We've got a little grain taking here you've managed a lot of the inputs you've got a good model you've 25 00:01:39,960 --> 00:01:46,620 matched the model to the problem and it's producing some some good outputs so you decide to report what 26 00:01:46,620 --> 00:01:47,220 you found. 27 00:01:48,150 --> 00:01:52,830 And after your initial report your Project Manager asked to see if the model can be improved to get 28 00:01:52,830 --> 00:01:53,660 better results. 29 00:01:54,390 --> 00:01:59,850 So you do your research and find there's another approach you can take. 30 00:02:00,060 --> 00:02:02,320 This is step six experiment. 31 00:02:04,320 --> 00:02:06,230 You might try a different model. 32 00:02:06,270 --> 00:02:11,730 We've got model to see how it goes and it performs pretty well then. 33 00:02:12,010 --> 00:02:16,480 If that didn't work when you reported it it is like oh maybe you could do a little bit better. 34 00:02:16,540 --> 00:02:24,640 You might try vary the inputs slightly and change your desired outputs and try a new model again. 35 00:02:24,760 --> 00:02:30,250 Don't worry if all of this looks confusing we're going to have plenty of practice doing this with hands 36 00:02:30,250 --> 00:02:32,670 on projects throughout the course. 37 00:02:33,540 --> 00:02:39,300 Remember at the start how we said there were three major steps we were going to do one create a framework 38 00:02:39,810 --> 00:02:46,540 to match the framework to data science and machine learning tools and three learn by doing well. 39 00:02:46,590 --> 00:02:49,800 After going through the last couple elections you've just ticked off. 40 00:02:49,800 --> 00:02:50,700 Step one. 41 00:02:51,120 --> 00:02:56,750 We've now got a framework to use for the modeling section of machine learning projects. 42 00:02:56,930 --> 00:02:58,540 Beautiful lenses. 43 00:02:58,710 --> 00:03:00,320 Step one problem definition. 44 00:03:00,450 --> 00:03:01,510 Step two data. 45 00:03:01,560 --> 00:03:08,490 Step three evaluation step four features step forward modeling Step Six experimentation all of this 46 00:03:08,490 --> 00:03:10,880 is an iterative process. 47 00:03:10,890 --> 00:03:18,300 Next up we're going to learn what tools we can use to apply the framework to different projects. 48 00:03:18,300 --> 00:03:25,410 And of course to do all this we're going to be working hands on writing machine learning code and creating 49 00:03:25,530 --> 00:03:31,920 projects of our own and remember to check the resources section for anything extra you may need. 50 00:03:31,990 --> 00:03:34,300 You can find all of what we've talked about there.