Exercise 2: Prepare Data for Training¶

Objective¶

In this lab, we will:

Load the Processed Dataset
Define Features and Target Variable
Prepare the Data for Model Training

What you will learn

How to select numerical and categorical features for a regression task.
The importance of the train-test split to prevent data leakage.
How to persist processed splits for downstream model training.

MLOps Perspective

Feature engineering and data splitting are often sources of "training-serving skew." By explicitly defining features and saving splits, we create a reproducible artifact that ensures the model is evaluated on data it has never seen. Data preparation is not just about cleaning; it's about creating a reproducible data pipeline. In a professional environment, these steps would be encapsulated in a pipeline component (like a Kubeflow component), ensuring that the exact same transformations are applied to training and production data.

Prerequisites¶

Cleaned dataset saved in directory data/processed

Step 1: Find and Open the Jupyter Notebook¶

In directory "labs/01_beginner/bike_demand_forecasting", please open the second notebook, "02_data_preparation.ipynb", and follow the instructions below to complete the tasks.

Step 2: Load the Processed Data¶

In this task, we take the cleaned dataset from the last task (data/processed) and prepare it for model training.

Add the cleaned data for the first three months in the appropriate cell:

data_2011_01.csv
data_2011_02.csv
data_2011_03.csv

Step 3: Set the Categorical Features¶

We already have set the numerical features.

You only need to set these categorical features:

['season', 'holiday', 'workingday', 'weathersit']

Step 4: Split the DataSet into Training and Test DataSets¶

In order to split the dataset into training and testing datasets, data scientists usually use train_test_split function from sklearn.model_selection module.

Just add this line of code in the appropriate cell, without any changes:

train_test_split(X_input, y_input, test_size=0.3, random_state=42)

Step 5: Inspect the Split Data¶

Check the shape of your training and test datasets to confirm the split was successful:

Just paste this line in a cell under the same notebook and run it.

print(X_train.shape, X_test.shape)

(Optional) You can also print out the first 5 rows of both training and test data:

print(X_train.head(), X_test.head())

Summary¶

In this exercise, you:

Loaded the cleaned dataset from the previous exercise
Defined features and target variable for training
Split the data into training and test datasets
Verified the split data shapes