Classification involves categorizing items into

predefined classes.

It's about predicting a specific class label from

input data.

Bias: Ensure evaluation metrics do not

introduce bias.

Definition of Classification

Example: Classifying supermarket items as

fruits/vegetables or grocery.

Ethical Considerations in Model

Transparency: Explain how metrics work and Evaluation
produce results.
Ethical Concerns

What is Classification?
Visual representation of classification with 4
and 2 classes.

Accountability: Take responsibility for the

choice of metrics.

Examples of classification use cases include

house price prediction, credit card fraud
detection, and salary prediction.
Classification Use Case Example

Accuracy measures the ratio of correct

predictions to total predictions.

The confusion matrix is a tool to assess model

Formula: (TP + TN) / (TP + TN + FP + FN). Accuracy from Confusion Matrix accuracy.

Example: Accuracy of 0.767.

Classification and
Evaluation Metrics It presents actual vs. predicted values for two
or more classes.

Accuracy is suitable for balanced datasets.

Confusion Matrix

Each cell in the matrix shows the number of

predictions in a category.
It may not be appropriate for unbalanced
datasets.
Limitations of Accuracy

The matrix helps visualize the model's

performance.
Other metrics like precision, recall, and F1-
score are needed.

The matrix shows the relationship between

predicted and actual values.
Precision is the ratio of correctly classified
positive examples to the total number of
predicted positive examples.

It helps to identify correct and incorrect

Classification Metrics predictions.
Understanding Confusion Matrix

Accuracy, Precision, Recall, and F1-

Precision is useful when false positives are Score
costly.
Precision

The matrix is used to calculate various

performance metrics.

Example: Predicting good weather for a

satellite launch.

True Positive (TP): Correctly predicted positive

class.

Recall measures the model's ability to identify

true positives.

True Negative (TN): Correctly predicted

negative class.

It is used when false negatives are costly. Recall

True Positive, True Negative, False Positive,
False Negative

False Positive (FP): Incorrectly predicted

positive class.
Example: Covid-19 prediction classifier.

F1-Score combines precision and recall into a

single metric.
False Negative (FN): Incorrectly predicted
negative class.

It is used for unbalanced datasets where it is

difficult to decide whether FP or FN is more
important.
F1-Score

Formula: 2 x (Precision x Recall) / (Precision +

Recall).