Corrigendum to “Assessing the knot in a noose position by thyrohyoid and cervical spine fracture patterns in suicidal hangings using machine learning algorithms: A new insight into old dilemmas”, [vol. 357, April 2024, 111973] (Forensic Science International (2024) 357, (S0379073824000549), (10.1016/j.forsciint.2024.111973))

Abstract

The authors regret the second paragraph of the section 2.4.1. Machine learning algorithm models development and analyses (starting with: “The following ML models were assessed: Multilayer Perceptron Artificial Neural Network (MLP–ANN) [50], Decision Tree (DT) [51], K-Nearest Neighbors (k-NN) [52], and Naïve Bayes (NB)… and ending with: “…reported in the Results section.”) is misleading and incomplete, and requires a corrigendum to explain the exact machine learning models development, program use, and to enable results reproducibility. The paragraph should read as follows: “The following ML models were assessed: Multilayer Perceptron Artificial Neural Network (MLP–ANN) [50], Decision Tree (DT) [51], K-Nearest Neighbors (k-NN) [52], and Naïve Bayes (NB) [53,54]. The finetuning of the MLP-ANN algorithms hyperparameters was done in Matlab, in an evolutionary manner, by using the Genetic algorithm (GA) [55], that starts a model optimization from an initial guess of hyperparameters (initial population), which are used as inputs for the objective function (OF). The OF ensembles the model with respect to the current guess of hyperparameters and computers the model accuracy – which needs to be maximized [55–58]. The hyperparameters of MLP-ANN considered for GA optimization were similar as in our previous work [56]. MLP-ANN algorithms with similar, corresponding accuracies to previously mentioned, and DT, K-NN, and NB algorithms were then developed all by automatic or repeated manual hyperparameter adjustments in SPSS. The settings and performances of algorithms developed in SPSS are reported in the Results section.” Figure 3 should be interpreted in accordance with this. We are truly sorry and apologize for any inconvenience caused by the mistake. © 2024 Elsevier B.V.