Time anomaly detection in the duration of civil trials in Italian justice

Figures & data

Figure 1. ER Diagram of the Domain.

Figure 2. The original Dataset.

Table 1. Comparison of the selected Classifiers with the best parameters identified through Grid Search.

Classifier	Class	Precision	Recall	F1 Score	Accuracy
Logistic Regression	Normal	0.92	0.99	0.96	0.92
	Slow	0.60	0.13	0.21
Gaussian NB	Normal	0.77	0.00	0.01	0.09
	Slow	0.09	0.99	0.16
Decision Tree	Normal	0.93	0.97	0.95	0.90
	Slow	0.39	0.22	0.28
Linear SVC	Normal	0.91	1.00	0.95	0.91
	Slow	0.35	0.02	0.04

Figure 3. Confusion Matrices for the tested Classifiers. (a) Logit. (b) Gaussian NB. (c) Decision Tree and (d) Linear SVC.

Figure 4. Confusion Matrices for the tested Classifiers after the Random Oversampling. (a) Logit. (b) Gaussian NB. (c) Decision Tree and (d) Linear SVC.

Table 2. Comparison of different Classifiers after Resampling.

Classifier	Class	Precision	Recall	F1 Score	Accuracy
Logistic Regression	Normal	0.63	0.80	0.70	0.66
	Slow	0.60	0.13	0.21
Gaussian NB	Normal	0.59	0.89	0.71	0.64
	Slow	0.78	0.39	0.52
Decision Tree	Normal	0.83	0.76	0.79	0.80
	Slow	0.78	0.84	0.81
Linear SVC	Normal	0.61	0.86	0.71	0.66
	Slow	0.77	0.46	0.58

Figure 5. Cumulative Explained Variance.