Dear all,
do you know what is the correct subsample size (number of pixel) to carry out a confusion matrix? I am digitizing the true classes on an image to check the accuracy of the classifier. However, digitizing the whole image is too much time-consuming and, so far, I have chosen a subsamble representative of the whole area/image. How much should it be large respect to the original image?
Thank you very much
In the context of a confusion matrix, there is no specific "subsample size." The confusion matrix is a tool used in classification tasks to visualize the performance of a machine learning model. It displays the counts of True Positives (TP), True Negatives (TN), False Positives (FP), and False Negatives (FN) for different classes in a tabular format.
The confusion matrix is constructed based on the entire dataset used for evaluation. It considers all the instances in the dataset and categorizes them according to their actual and predicted class labels. Each entry in the confusion matrix represents the count of instances falling into a specific category, such as TP, TN, FP, or FN.
Here's a breakdown of the terms used in a confusion matrix:
- True Positives (TP): The number of instances correctly predicted as positive.
- True Negatives (TN): The number of instances correctly predicted as negative.
- False Positives (FP): The number of instances incorrectly predicted as positive when they are actually negative (Type I error).
- False Negatives (FN): The number of instances incorrectly predicted as negative when they are actually positive (Type II error).
The confusion matrix is typically constructed using the entire dataset, not a subsample of it. It provides valuable insights into the model's performance across all classes and helps in evaluating its accuracy, precision, recall, F1-score, and other metrics.
Thank you very much for your precisation. I then also found this paper demonstrating that it is sufficient a very smaller area: Blatchford, M. L., Mannaerts, C. M., & Zeng, Y. (2021). Determining representative sample size for validation of continuous, large continental remote sensing data. International Journal of Applied Earth Observation and Geoinformation, 94, 102235.
- Badges
- Science topic
- Similar topics
- Organizational Psychology
- Training