Make Predictions and Analyze Results

Quick reference

• 00:05 Okay, so let's make some predictions and analyze some results.
• 00:09 So let's create a variable called predictions.
• 00:12 And let's set that equal to, now we can call lin.predict.
• 00:17 And we can just pass in what do we want to predict?
• 00:20 Well, we want to predict on X_test,
• 00:23 not our trained to data but our test data.
• 00:29 And we can run, we can just see that what this is.
• 00:34 And we get an array with a bunch of price data, right?
• 00:38 So it doesn't mean a whole lot to us, but
• 00:40 we can make a scatterplot to kind of look at this.
• 00:42 So let's go plt.scatter, and we want to pass in y_test,
• 00:50 and then our X_test, predictions, so we can just add in predictions.
• 00:56 When we do that, we get this scatterplot, and
• 00:59 you'll notice it's sort of seems linear, right?
• 01:02 Not the most linear, when we looked at this linear regression
• 01:06 on Wikipedia these were really kind of tightly bunched around this line.
• 01:11 Whereas these, if we pictured a line going through here,
• 01:15 these aren't really tightly bunched around,
• 01:17 there's a fit line going through here, there's a bunch of other outlier.
• 01:22 So this may or may not be the best model to use linear regression.
• 01:27 And we can sort of see, in order for a linear regression to really work well,
• 01:33 that data needs to be normally distributed.
• 01:36 So we can sort of use seaborn real quick to just check whether or
• 01:41 not this is normally distributed by calling a distplot, and
• 01:46 we can pass in our y_test- predictions.
• 01:52 And when we do that, okay, it looks sort of normally distributed,
• 01:56 but we've got this tail over here that's definitely not normal.
• 02:00 Maybe account for this stuff over here that's sort of not in line with everything
• 02:04 else here, so there may be some outliers in our data, super expensive houses,
• 02:10 maybe react differently to market data, who knows?
• 02:14 It's really interesting to look at.
• 02:18 If we run this for our training data, Which we really wouldn't want to do.
• 02:25 But this looks a little more normally distributed.
• 02:33 But not really when we look at the chart.
• 02:37 So that's our linear regression model.
• 02:40 Like I said, very, very quick overview for this, just to kind of whet
• 02:45 your appetite and show you sort of things that are available for future learning.
• 02:49 Now that you've learned an introduction to data analysis with Python,
• 02:53 machine learning is the next thing you're going to want to focus on.
• 02:56 And this is just a taste of machine learning.
• 02:59 So just a quick recap, we loaded our Boston data,
• 03:02 we slapped it into a data frame, we set our x and
• 03:05 our y as the columns of that data frame, the y is the actual price data.
• 03:11 And we split our data into training and testing groups.
• 03:14 We ran our linear regression, we fit our model to our x train and y train.
• 03:21 We found our intercepts and our coefficients.
• 03:23 Coefficients give us data that relates to a one unit increase and
• 03:28 x correlates to an increase of whatever this thing is in prices, which may or
• 03:33 may not make sense if you go through here.
• 03:35 But remember, this is very, very old data that may not reflect reality very well,
• 03:40 and sort of keep that in mind.
• 03:42 We can run predictions, we can create predictions, and then analyze our results.
• 03:48 So that's all for this video.

Lesson notes are only available for subscribers.