A classification system is expected to classify all data sets correctly, but it can not be denied that the performance of a system cannot be 100% perfect so that a classification system must also be measured for performance. Generally, classification performance measurements are performed with a confusion matrix. Configuration matrix is a table of work of classification.
By knowing the amount of data classified correctly, we can know the accuracy of predicted results, and by knowing the amount of data classified correctly, we can know the accuracy of predicted results, and by knowing the amount of data classified wrongly, we can know the error rate of prediction which is conducted. These two quantities are used as the matrix performance classification. Here's a formula to calculate accuracy.
Accuracy = Amount of predicted data correctly / number of predictions made.
To calculate the error rate the formula is used.
Error rate = an Incorrect number of predicted data / Number of predictions made.
All classification algorithms attempt to form models that have high accuracy. Generally, the built model can predict correctly on all data into the training data, but when the model is dealing with the test data, then the model performance of a classification algorithm is determined. MATLAB provides a function to compute the confusion matrix from the predicted class of results compared to the actual class. With syntax C = Confusionmat (group, grouphat). Parameter C is a square matrix of length = width of class number. The sum of all its elements is equal to the number of group vector and grouphat elements. C (i, j) is the number of observations from the mapping results of group (i) to grouphat (j). The Group is a vector (1 column matrix) containing the original class of test data and Grouphat is a vector (1 column matrix) containing the predicted class.
By knowing the amount of data classified correctly, we can know the accuracy of predicted results, and by knowing the amount of data classified correctly, we can know the accuracy of predicted results, and by knowing the amount of data classified wrongly, we can know the error rate of prediction which is conducted. These two quantities are used as the matrix performance classification. Here's a formula to calculate accuracy.
Accuracy = Amount of predicted data correctly / number of predictions made.
To calculate the error rate the formula is used.
Error rate = an Incorrect number of predicted data / Number of predictions made.
All classification algorithms attempt to form models that have high accuracy. Generally, the built model can predict correctly on all data into the training data, but when the model is dealing with the test data, then the model performance of a classification algorithm is determined. MATLAB provides a function to compute the confusion matrix from the predicted class of results compared to the actual class. With syntax C = Confusionmat (group, grouphat). Parameter C is a square matrix of length = width of class number. The sum of all its elements is equal to the number of group vector and grouphat elements. C (i, j) is the number of observations from the mapping results of group (i) to grouphat (j). The Group is a vector (1 column matrix) containing the original class of test data and Grouphat is a vector (1 column matrix) containing the predicted class.
Advertisement
No comments