Binary-classification-metrics

In this repository we build a binary classification model using Logestic Regression, and use different binary classification metrics to have a better understanding of them.

How to use

Dataset:

• https://www.kaggle.com/datasets/mlg-ulb/creditcardfraud

• Go to your kaggle account --> Your profile --> Settings --> API --> Create New Token.

A json file (kaggle.json) will be downloaded automatically. Upload it in colab/files

Binary Classification Metrics

• confusion matrix: A confusion matrix is a tabular summary that helps evaluate the performance of a classification model. Let's break it down:

• It summarizes the number of correct and incorrect predictions made by the model.

• The matrix is especially useful for assessing classification models beyond basic accuracy metrics.

• Here's how it works:

  • True Positives (TP): These occur when the model accurately predicts a positive data point.
  • True Negatives (TN): These occur when the model accurately predicts a negative data point.
  • False Positives (FP): These occur when the model predicts a positive data point incorrectly.
  • False Negatives (FN): These occur when the model mispredicts a negative data point.

Using confusion matrix we can calculate the following performance metrics.

Most Common Metrics:

• Accuracy: The ratio of total correct instances to the total instances.

  $Accuracy$ = $\frac{TP + TN}{TP + TN + FP + FN}$

• Precision: The proportion of true positive predictions among all positive predictions.

  $Precision$ = $Positive$ $Predictive$ $Value$ = $\frac{TP}{TP + FP}$

• Recall (Sensitivity): The proportion of true positive predictions among all actual positive instances.

  $Recall$ = $Sensitivity$ = $True$ $Positive$ $Rate$ = $\frac{TP}{TP + FN}$

• F1-score: A harmonic mean of precision and recall.

  $F1$ = $\frac{2 \times Precisicion \times Recall}{TP + FN}$

Less Common Metrics:

• $Specificity$ = $True$ $Negative$ $Rate$ = $\frac{TN}{TN + FP}$

• $False$ $Negative$ $Rate$ = $\frac{FN}{TP + FN}$

• $False$ $Positive$ $Rate$ = $\frac{FP}{TN + FP}$

• $Negative$ $Predictive$ $Value$ = $\frac{TN}{TN + FN}$

• $Balanced$ $Accuracy$ = $\frac{(TPR + TNR)}{2}$

• $Matthews$ $Correlation$ $Coefficient$ = $\frac{(TP * TN) - (FP * FN)}{\sqrt{(TP + FP) * (TP + FN) * (TN + FP) * (TN + FN)}}$

Confusion matrix is a powerful tool for understanding how well your model is performing!

Logestic Regression

Logistic regression is a powerful technique used in machine learning for classification tasks.

1. What is Logistic Regression?

   • https://www.youtube.com/watch?v=yIYKR4sgzI8
   • Logistic regression predicts the output of a categorical dependent variable (e.g., Yes/No, 0/1, True/False).
   • Instead of giving exact values like 0 or 1, it provides probabilistic values between 0 and 1.
   • It's an extension of linear regression, mainly used for classification problems.
   • The goal is to predict the probability that an instance belongs to a given class.

2. Logistic Function (Sigmoid Function):

• The sigmoid function maps real values to probabilities within the range of 0 and 1.
• It forms an "S" shaped curve and ensures the output stays within the valid probability bounds.
• The threshold value (usually 0.5) determines whether an instance belongs to Class 0 or Class 1.

3. Types of Logistic Regression:
   • Binomial: Only two possible dependent variable types (e.g., Pass/Fail).
   • Multinomial: Three or more possible unordered dependent variable types (e.g., "cat," "dogs," "sheep").
   • Ordinal: Three or more possible ordered dependent variable types (e.g., "low," "Medium," "High").