Accuracy Evaluation is an evaluation of the performance of prediction models created from prediction model creation (training) data. Easy Predictive Analytics randomly extracts a portion of the prediction model creation (training) data, creates a prediction model using the remaining data, and makes a prediction on the data extracted earlier. By comparing the prediction with the correct answer, we evaluate how much we can predict the data that we have never seen before.

You can also specify the data to use for accuracy evaluation by specifying the file from “Specify Data for Evaluation”.

Accuracy evaluation metrics depends on the problem setting. See the links below for more information on each metric.

Evaluation Metrics for Binary Classification

• Tips [How to read prediction accuracy (binary classification)]
• AUC
• Accuracy
• Precision
• Recall
• F-score

Evaluation Metrics for Multiclass Classification

• Tips [How to read prediction accuracy (multiclass classification)]
• Accuracy
• Precision
• Recall
• F-score

Evaluation Metrics for Regression

• Tips [How to read prediction accuracy (regression)]
• Mean Absolute Error
• RMSE
• Median Absolute Error
• Median Absolute Error Ratio
• Coefficient of Determination