Mode Choice Analysis Using Random Forrest Decision Trees

Unsupervised Anomaly Detection of Healthcare Providers Using Generative Adversarial Networks

“…Classification Algorithms: Four binary classification algorithms were applied in the second part of the study to give interpretability to the anomalies. The algorithms applied include the logistic regression (LR), extreme gradient boosting (XGB), random forest (RF) and decision tree (DT) [22]. These algorithms are summarised highlighting their core capability.…”

“…Random Forest (RF) is a tree constructed algorithm from a set of possible trees with random features at each node. Random forest can be generated efficiently and the combination of large sets of random trees generally leads to accurate models to detect anomalies [22]. Moreover, the random forest algorithm has been used in this study due to its versatility in being applied to large data sets and feature importance [6].…”

“…In the current context of anomaly classification, the decision tree model outputs a predicted target class (anomaly or normal) for each terminal node produced. Decision Trees automatically reduce complexity, selection of features and the predictive analysis structure is understandable and interpretable [22].…”

“…The LR, XGB, RF and DT algorithms are successful in detecting anomalies [6,18,22] however their main use in the current study is their ability to generalize, feature selection, interpretability [6,18,22,30] and further explain how the features contribute to the anomalous healthcare providers. This explanation was further achieved through the use of SHapley Additive exPlanation (SHAP) [13].…”

Learning Condition–Action Rules for Personalised Journey Recommendations

“…Omrani [11] find that neural network-based approaches (multi layer perceptron and radial basis function networks) have higher performance than multinomial logistic regression and support vector machines. Sekhar et al [13] find that the random forest classifier out-performs the multinomial logit model. Hagenauer et al [7] compare seven classifiers applied to the task of predicting travel mode based on a number of inputs.…”