Through the use of confusion matrices, this project evaluates and analyzes various classification models by comparing their accuracy, precision, sensitivity, and specificity statistics for categorizing the test times data by sort type for bubble and selection sorts.

The data contains two qualitative columns of which one was converted into a dummy variable beforehand. Then, since bubble and insertion sort performed similarly with different ordered data in the analysis, the two of them were grouped together, where group zero represents bubble and insertion sort and group one represents selection sort. The dependent variable was stored in a Pandas series while the independent variables were stored in a data frame as required by Scikit-learn. They were then split into two datasets consisting of 70% training and 30% testing data. The source code can found on Github.

The confusion matrices are evaluated by comparing the accuracy, precision, sensitivity, and specificity for each model.

• Accuracy refers to the frequency of correct predictions. The ratio of the true cases to all the cases.

• Precision refers to the frequency of positive events that turned out to be true (positive).

• Sensitivity refers to the ratio of correct positive events to total positive events.

• Specificity refers to the ratio of correct negative events to total negative events.

For our data, this means the following: 

• If the observations and predictions are both zero, the model correctly classified bubble sort as bubble sort (or true negative).

• If the observations is zero and the predictions is one, the model incorrectly classified bubble sort as selection sort (or false positive).

• If the observations is one and the predictions is zero, the model incorrectly classified selection sort as bubble sort (or false negative).

• If the observations and predictions are both one, the model correctly classified selection sort as selection sort (or true positive).