Open a terminal (CMD or PowerShell) in the extraction directory and execute: confusion.exe Use code with caution. Copied to clipboard
To generate a report using the Windows version, follow these steps:
For multi-class datasets, the report provides both macro-averaged (equal weight to each class) and micro-averaged (equal weight to each instance) scores. Confusion-0.5-win.zip
Confusion is a command-line utility designed for data scientists and researchers. Its primary purpose is to take a set of predicted labels and actual (ground truth) labels to produce a detailed breakdown of a model's performance beyond simple accuracy. 0.5 Platform: Windows (Intel/AMD 64-bit)
By highlighting high values in non-diagonal cells, the tool helps identify which specific classes the model is frequently "confusing" with others. Open a terminal (CMD or PowerShell) in the
Unzip Confusion-0.5-win.zip to a local folder.
Below is a detailed report on the software, its core functionality, and how to utilize it for performance evaluation in classification tasks. Its primary purpose is to take a set
The tool processes classification data to build a matrix where rows represent actual classes and columns represent predicted classes. Definition Importance Correctly predicted positive instances. Measures successful identification. False Positives (FP) Incorrectly predicted positive instances. Indicates "Type I" error (false alarm). False Negatives (FN) Missed positive instances. Indicates "Type II" error (missed detection). Precision Shows how reliable the positive predictions are. Recall (Sensitivity) Shows how many actual positives the model found. Usage Instructions