Neural Network Tutorials - Herong's Tutorial Examples

∟Performance Evaluation Metrics

∟CI (Concordance Index)

This section describes CI (Concordance Index) as a metric to evaluate the performance of a continuous prediction model.

What Is CI (Concordance Index)? - CI is a commonly used metric to evaluate the performance of a continuous prediction model.

Given a prediction model and a set of test samples, the CI of the model on the test set is defined below:

where a sample pair (y₁, y₂) is a concordant pair, if and only if (y₁ - y₂) has the same sign as (y'₁, y'₂).

CI can also be expressed as:

where h(x) is a step function:

CI measures how well the predicted value set matches the actual value set in terms of a sorted order. For example, comparing the sorted sequence of predicted values with the sorted sequence of actual values:

• If CI = 1, all samples are in the same order in both sequences. This indicates a good prediction.
• If CI = 0.5, half of samples are in the correct order and half of samples are in the wrong order. This indicates a poor prediction. Actually, a random prediction will result CI = 0.5.
• If CI = 0, all samples are in the reverse order in both sequences. This indicates that something went wrong in the prediction.

Table of Contents

 About This Book

 Deep Playground for Classical Neural Networks

 Building Neural Networks with Python

 Simple Example of Neural Networks

 TensorFlow - Machine Learning Platform

 PyTorch - Machine Learning Platform

 Gradio - ML Demo Platform

 CNN (Convolutional Neural Network)

 RNN (Recurrent Neural Network)

 GNN (Graph Neural Network)

 GAN (Generative Adversarial Network)

►Performance Evaluation Metrics

 MSE (Mean Squared Error)

►CI (Concordance Index)

 PCC (Pearson Correlation Coefficient)

 References

 Full Version in PDF/EPUB

CI (Concordance Index) - Updated in 2023, by Herong Yang