Künstliche Intelligenz und Maschinelles Lernen in der Praxis

1. 00:00After we have just had a good image about our data, we can
2. 00:04now in this unit finally build our first AI model.
3. 00:08And the first thing we do is, what do we do? is the data we just need to
4. 00:14to bring into an AI format.
5. 00:16So let's look at our data again then we see our different attributes,
6. 00:21Longitude, Latitude and even also the Ocean Proximity.
7. 00:27Now we just need to know that many AI models can only work on numerical data,
8. 00:34for example, only count on numbers. What is relatively common, considering that
9. 00:39many of these models simply formulas exist and one is relatively difficult
10. 00:44categories.
11. 00:46That is why we must now consider how we get the data in a suitable AI format.
12. 00:52But that is actually relatively simple. What we can do for it is
13. 00:56one-hot encodings must be calculated, also one-hot transformations.
14. 01:01And that's relatively simple. by simply saying that we
15. 01:05Transform categories into binary vectors.
16. 01:09That means, for example, that we have a category close to the sea, then from this category
17. 01:16in our data a column where we say category close to the sea is true or not true.
18. 01:24So make it quasi a binary attribute.
19. 01:28It's easy here too.
20. 01:29because it's such a typical transformation. We don't have to reinvent this completely.
21. 01:35You can also again use pandas for.
22. 01:37So we notice pandas offers really a variety of features that we keep
23. 01:42and the already for us.
24. 01:45Why it's so easy in part is popular, just how easy it is for us
25. 01:49reuse.
26. 01:50And this function that we have here from Pandas , is called get dummies, meaning dummy values.
27. 01:56That's exactly what we can do with our data exactly transform into the one-hot encoding.
28. 02:02It's also always there It was incredibly fast.
29. 02:07Right, let's just look at it Let me show you what it looks like now.
30. 02:10So we had the attribute ocean before Proximity, and we had four more
31. 02:17expressions, namely less than one hour Ocean, Inland, near Bay and near Ocean.
32. 02:23And now we don't have any more Category, but we now have four attributes.
33. 02:27And they're just represented as binary.
34. 02:30So 1 stands for yes, the value comes before and 0 means no, it doesn't appear.
35. 02:37Exactly, our AI model then work well later.
36. 02:43What's still going to have to do is make this data Frame, that is to prepare our data structure in this way,
37. 02:49or to serve us in this way that we also enter input data and target data
38. 02:55So we want our model. training on something so that it can learn well.
39. 03:00We can do this very simply by again a very simple help function
40. 03:04, which are ultimately from our dataframe into the input matrix, that's what we know.
41. 03:12So that's where we put that. Want to train the model.
42. 03:14And that's all up to on the median house value because
43. 03:17That is exactly what we want to predict. And as a target vector, that is, what we call the model
44. 03:25, but what it will not be at runtime , but only for the training data
45. 03:29are the actual median house values, So, what the model learns is our labels.
46. 03:37And we define this function for ourselves.
47. 03:39then on our data.
48. 03:43Yes, so now have input matrix and target vectors for both training failures and test cases.
49. 03:51And we are now ready to finally build our first AI model.
50. 03:55And that's what we're doing now.
51. 03:58And more than that It's actually not at all.
52. 04:01That is, when we do the execution, we have right now our first AI model
53. 04:09built for this course and already trained.
54. 04:11So it went really, really easy. we can use this model now
55. 04:18for forecasts. Of course, you can also I'm going to do a bunch of other stuff here.
56. 04:22So there are other models or hyper parameters, but in and of itself that would be
57. 04:28for example, a useful Machine learning model.
58. 04:31What we simply emphasize again just want it to be in such a
59. 04:36AI project often simply stops at it It's important to understand the data well.
60. 04:39Even if it is somehow a highly modern is an AI algorithm, i.e.
61. 04:43artificial neural network, you should Actually never just without the data
62. 04:48to have looked at this model and train on it and ultimately this
63. 04:55Training is often done relatively quickly.
64. 04:56But just when you know what the data is for example, you have an approximate
65. 05:01understanding, usually you are much better can actually work with the models.
66. 05:05And we'll also have different View other models in the course.
67. 05:09So don't worry, we'll still The whole bandwidth is looking at us.
68. 05:14But now for this first project we have We just decided to do it.
69. 05:18simple model to use and look at us then add more models later.
70. 05:23Right, and now we want to move on to the next Watch video once, how good this model then
71. 05:29finally performs and whether we do want to use or theoretically still further
72. 05:35on our data.
73. 05:36So let's look at that in the next video.

To enable the transcript, please select a language in the video player settings menu.