Machine Learning: Classification and Logistic Regression

Vocabulary flashcards covering core concepts of supervised machine learning, distinguishing regression from classification, and implementing binary logistic regression.

Classification

A type of supervised machine learning that predicts a category or class rather than a numerical value based on patterns learned from labeled training examples.

Regression

A type of machine learning prediction task designed to predict continuous numerical values, such as student marks or house prices.

Features ($x$)

The input columns or relevant pieces of information provided to a model, such as study hours and attendance, denoted by $x$.

Target ($y$)

The outcome or value that the machine learning model is trained to predict, denoted by $y$, such as pass or fail status.

Training Set

The portion of the dataset (typically 80%) used by the model to learn coefficients, patterns, and relationships.

Testing Set

A separate portion of the dataset kept aside to evaluate how well the trained model performs on unseen examples.

Binary Classification

A classification problem with exactly two possible classes or categories, such as pass/fail or spam/not spam.

Multiclass Classification

A classification problem with more than two possible classes, such as categorizing types of flowers or species of animals.

Logistic Regression

An algorithm commonly used for classification that estimates the probability that an observation belongs to a particular class and uses a threshold to assign a class label.

Sigmoid Function

A mathematical function that transforms any real-valued score calculated by the model into an estimated probability value between $0$ and $1$.

Decision Threshold

The cutoff value (commonly set to a default of $0.50$) used to convert an estimated probability into a final predicted class label.

Scikit-learn

A Python library that provides tools and traditional machine learning algorithms for tasks such as data splitting, model training, and evaluation.

Accuracy Score

An evaluation metric that calculates the proportion of correct predictions made by the model relative to the total number of test examples.

Confusion Matrix

A table that compares actual target classes with predicted classes, showing true negative, false positive, false negative, and true positive results.

predict_proba

A Scikit-learn method that returns the model's estimated probabilities for each possible class rather than only the final categorical prediction.