Introduction to Supervised Machine Learning

1. Regression: Predicting Continuous Values

Regression is used when the target output is a continuous, numerical value. In these scenarios, we are not asking "Is this A or B?", but rather "How much?" or "How many?". Common examples include predicting the price of a house, the temperature for tomorrow, or the future value of a stock.

For instance, in a simple linear regression model, we might try to predict a house price based on its square footage. The relationship can be expressed as:

$$\begin{aligned} y &= w \cdot x + b \end{aligned}$$

Where $w$ represents the weight (slope) and $b$ represents the bias (intercept). The model learns the optimal values for $w$ and $b$ to fit the data points as closely as possible.

2. Classification: Predicting Discrete Categories

Classification is used when the target output is a discrete label or a category. Instead of predicting a number on a continuous scale, the model assigns the input to one of several predefined classes.

Classification can be divided into two sub-types:

• Binary Classification: There are only two possible outcomes (e.g., Spam vs. Not Spam, Malignant vs. Benign).
• Multi-class Classification: There are more than two possible outcomes (e.g., classifying an image as a "dog," "cat," or "bird").

A common algorithm for classification is the Logistic Regression model (despite its name, it is used for classification). It uses a sigmoid function to map any real-valued number into a probability between 0 and 1, which is then used to decide which class the input belongs to.

Summary Table: Regression vs. Classification

To help solidify these concepts, refer to the following comparison:

The Workflow of Supervised Learning

To implement a supervised learning model effectively, a practitioner typically follows these steps:

1. Data Collection: Gathering raw data that contains both features and labels.
2. Data Preprocessing: Cleaning the data, handling missing values, and scaling features so they are on a similar magnitude.
3. Splitting the Dataset: Dividing the data into a Training Set (used to teach the model) and a Test Set (used to evaluate how well the model performs on data it hasn't seen before).
4. Model Training: Feeding the training data into the algorithm to minimize the loss function.
5. Evaluation: Using the test set to calculate metrics like Mean Squared Error (for regression) or Accuracy/Precision/Recall (for classification).

By following this structured approach, we ensure that the model isn't just "memorizing" the training data (a problem known as overfitting), but is actually learning the underlying patterns that allow it to generalize to the real world.