Metrics.jl

This package is a collection of diverse Metrics to analyse the performance of variety of Machine Learning and Deep Learning Models. These include Regression Models, NLP Models, CV Models, Recommendation Models and further utilities to support better user interface.

Installation

To install Metrics.jl, you need to fill in the following code into the Julia Prompt:

] add Metrics

Basic Usage

Suppose you are working on a Classification Problem. Once you have your model ready, and you would like to evaluate your model's performance. One of the most obvious way would be to use the magnitude of loss function as evaluation metrics. However, a better way to accomplish this could be to use Metrics.jl, using the following commands:

using Metrics

Metrics.report_stats(y_pred, y_true)  # where y_pred are the predicted values and y_true are onehot_encoded ground truth values.

This will print the performance statistics of the model, based on the provided y_pred and y_true values. This statistics include the Confusion Matrix, Accuracy, Precision, Recall, F1 Score and much more. Where Metrics.jl provide you option to get complete evaluation of the model using multiple statistics functions within a single function, it also provide you option to use these statistics functions individually as per your choice. Their usage, you can find further in this documentation.