One of the metrics that assess the predictive performance of a model, which is the percentage of the total number of accurate predictions that the model has made Structured Data Categories Classification is as follows.
The definitions for each term(TP, TN, FP, FN) can be found in more detail at Confusion Matrix .
Accuracy is defined slightly differently for a continuous value classification(Regression) model. Unlike category classification models that only predict binary results (True/False), continuous value classification models find it difficult to predict the correct answer perfectly. Therefore, the focus is on how close the forecasts are to the answers, not on whether the criteria for predictive accuracy are correct. In other words, the smaller the error between the model's forecast and the correct answer, the more accurate the model is. The most common way to calculate the error is root mean squared error(RMSE). Suppose the actual value of sample i is , RMSE equation is as follows.
- Accuracy Calculation of a Categories Classification Model
The table above is an example of predicted values and actual correct answer values of the category classification model. TP = 1, FN = 2, TN = 1, FP = 1. The accuracy for this is calculated as follows.
- Accuracy Calculation of a Regression Classification Model
The table above shows the example of predicted values and the actual correct answers for the continuous classification model. The residuals(predicted-actual) are 1, -3, 1, 1, and -1, respectively. Using these values, RMSE is calculated as follows
[Click a tab 'Details' of the model you want to check. Select Model 27 whose accuracy is the highest.]
[You can check the accuracy of the model through the accuracy bar graph at the bottom left.]