Tabular Data and Terminology

Terminology

Here is some basic terminology used in ML:

• examples = rows
• features = inputs
• targets = outputs
• training = learning = fitting

{fig-alt:“Supervised machine learning terminology” fig-align=“center” width=“90%”}

In the supervised machine learning paradigm, we have input data and an output. We feed our input to a machine learning algorithm.

The question is how do we effectively represent this input?

Is there a specific required format for our data so that we can pass it to machine learning algorithms.

• YES! In supervised machine learning, we typically work with tabular data.

Here is a toy example of tabular data.

The task here is to predict the quiz2 grade given all this information.

• Rows are examples
• Columns are features and one of the columns is typically the target.
• Features are relevant characteristics of the problem (usually suggested by experts).
• To a machine, column names (features) have no meaning. Only feature values and how they vary across examples mean something.
• Training a model can also be called learning or fitting a model.

All of these will be used in the course so it’s important to get familiar with the vocabulary now.

You will see a lot of variable terminology in machine learning and statistics and sometimes they can be confusing. See the MDS terminology resource here to clear up any confusions.

Terminology

Example 1: Tabular data for the housing price prediction problem

df = pd.read_csv("data/kc_house_data.csv")
df.head(3)

	price	bedrooms	bathrooms	sqft_living	...	lat	long	sqft_living15	sqft_lot15
0	425000.0	3	1.00	1180	...	47.7121	-122.333	2240	7875
1	230500.0	2	1.00	740	...	47.4793	-122.336	850	8775
2	372000.0	4	2.75	2610	...	47.3747	-122.124	2390	7852

3 rows × 19 columns

df.shape

(1000, 19)

Let’s look at some examples of terminology:

Let’s go back to our housing price prediction problems. In this particular example, there are 18 features and 21613 examples.

Our target column is the price column.

Example 2: Tabular data for quiz2 classification problem

classification_df = pd.read_csv("data/quiz2-grade-toy-classification.csv")
classification_df.head(3)

	ml_experience	class_attendance	lab1	lab2	lab3	lab4	quiz1	quiz2
0	1	1	92	93	84	91	92	A+
1	1	0	94	90	80	83	91	not A+
2	0	0	78	85	83	80	80	not A+

classification_df.shape

(21, 8)

Here is another example.

This is our toy quiz2 classification data. The target here is quiz2. Our task is to predict whether a student gets an A+ or not A+ in quiz2.

Since this is a toy dataset. We only have 21 examples and 7 different features with our target column being quiz2.

X = classification_df.drop(columns=["quiz2"])
y = classification_df["quiz2"]
X.head()

	ml_experience	class_attendance	lab1	lab2	lab3	lab4	quiz1
0	1	1	92	93	84	91	92
1	1	0	94	90	80	83	91
2	0	0	78	85	83	80	80
3	0	1	91	94	92	91	89
4	0	1	77	83	90	92	85

y.head()

0        A+
1    not A+
2    not A+
3        A+
4        A+
Name: quiz2, dtype: object

In order to train a model, we need to separate our data into features and the target.

We save our features which are columns ml_experience to quiz1 in an object named X.

Our target column is quiz2 and gets saved in an object named y.

We will explain why we do this in the next set of slides.

Let’s apply what we learned!