Precision, Recall and F1-Score using R

In machine learning, evaluating model performance is critical. Three widely used metrics—Precision, Recall, and F1-Score—help assess the quality of classification models. Here's what each metric represents:

1. Recall: Measures the proportion of actual positive cases correctly identified. Also known as sensitivity or true positive rate (TPR).
   Formula: \text{Recall} = \frac{\text{True Positives (TP)}}{\text{True Positives (TP)} + \text{False Negatives (FN)}}
2. Precision: Represents the proportion of positive predictions that are correct.
   Formula: \text{Precision} = \frac{\text{True Positives (TP)}}{\text{True Positives (TP)} + \text{False Positives (FP)}}
3. F1-Score: The harmonic mean of Precision and Recall, offering a balance between the two.
   Formula: \text{F1-Score} = 2 \times \frac{\text{Precision} \times \text{Recall}}{\text{Precision} + \text{Recall}}

Calculating Precision, Recall and F1 Score in R

R provides robust tools to compute these metrics easily. Below are step-by-step instructions using caret and Metrics packages.

Step 1: Install and Load Required Packages

To compute Recall and related metrics, first install the necessary R packages.

install.packages("caret")
install.packages("Metrics")

Load the packages:

library(caret)
library(Metrics)

Step 2: Compute Recall Using the caret Package

The caret::confusionMatrix() function computes a confusion matrix along with Recall and other metrics. Here's an example:

# Create predicted and actual class labels
predicted <- c(1, 1, 1, 0, 0)
actual <- c(1, 0, 1, 1, 1)

# Generate confusion matrix
xtab <- table(predicted, actual)
cm <- caret::confusionMatrix(xtab)
print(cm)

Output

Confusion Matrix and Statistics

         actual
predicted 0 1
        0 0 2
        1 1 2

               Accuracy : 0.4             
                 95% CI : (0.0527, 0.8534)
    No Information Rate : 0.8             
    P-Value [Acc > NIR] : 0.9933          

                  Kappa : -0.3636         

 Mcnemar's Test P-Value : 1.0000          

            Sensitivity : 0.0000          
            Specificity : 0.5000          
         Pos Pred Value : 0.0000          
         Neg Pred Value : 0.6667          
             Prevalence : 0.2000          
         Detection Rate : 0.0000          
   Detection Prevalence : 0.4000          
      Balanced Accuracy : 0.2500          

       'Positive' Class : 0               

Step 3: Compute Metrics Using the Metrics Package

The Metrics package offers functions like recall(), precision(), and f1_score() for quick calculation.

# Install the necessary packages if not installed
install.packages("Metrics")
library(Metrics)

# Predicted and actual labels
predicted <- c(1, 1, 1, 0, 0)
actual <- c(1, 0, 1, 1, 1)

# Compute Recall (correct order: actual, predicted)
recall_score <- recall(actual, predicted)

# Compute Precision (correct order: actual, predicted)
precision_score <- precision(actual, predicted)

# Compute F1-Score (correct order: actual, predicted)
f1_score <- f1(actual, predicted)

# Print results
cat("Recall:", recall_score, "\n")
cat("Precision:", precision_score, "\n")
cat("F1-Score:", f1_score, "\n")

Output

Recall: 0.5 
Precision: 0.6666667 
F1-Score: 1 

• Recall is for applications like disease detection or fraud detection where missing positive cases is costly.
• Precision is for spam filters or recommendation systems where false positives should be minimized.
• F1-Score balances precision and recall, making it ideal for imbalanced datasets.

R programming makes it simple to compute metrics like Precision, Recall, and F1-Score using packages like caret and Metrics. These metrics are indispensable for evaluating and improving model performance, particularly for binary classification problems.