Charles J. Stone’s research while affiliated with Belmont College and other places

... Our evaluation process involved four machine learning models -Logistic Regression with elastic net penalization [35,36], Decision Tree [36,37], Random Forest [36,38], and XGBoost [39]. We chose these classical machinelearning algorithms due to the small amount of available data (229 participants) and their reputation to work well even in such scenarios. ...

Reference:

RADAR-AD: assessment of multiple remote monitoring technologies for early detection of Alzheimer’s disease

... On the other hand, the different condition ( ( , )/| |) represents the belonging to class . By separating different probabilities in the formula, different situations are evaluated in more detail (Breiman et al. 1984;Pal 2005). ...

Reference:

Monitoring the rehabilitation process of the windthrow area using UAS images and performance comparison of Sentinel-2A based different vegetation indexes

... To predict the climatic match, six methods were fitted to determine the model performance and relate the predictor variable that best described the respective avian species distribution suitability. The methods were isolated from 15 methods supported by the SDM package with the following methods; boosted regression tree (BRT: Friedman 2001), classification and regression trees (CART : Breiman 1984), generalised linear model (GLM: McCullagh 1989), multivariate adaptive regression spline (MARS: Friedman 1991), random forest (RF : Breiman 2001), and support vector machine (SVM: Vapnik 1995). The models were evaluated at 100 runs of bootstrap replication to give model adequate time to converge. ...

Reference:

Impact assessment of seven alien invasive bird species already introduced to South Africa

... The machine learning algorithms used in the experiment were Support Vector Classification (SVC) [49], k-Nearest Neighbor (k-NN) [50], Decision Tree (DT) [51], Random Forest (RF) [52], Artificial Neural Network (ANN) [53], Gradient Boosting Decision Tree (GBDT) [54], and TabNet [55]. These algorithms except for TabNet were implemented by scikit-learn [56], and the TabNet algorithm was implemented by the pytorch-tabnet library [57]. ...

Reference:

Self-Adaptive Traffic Anomaly Detection System for IoT Smart Home Environments

... We tested which subsets of the 17 variables were most strongly associated with the establishment and impact stages of invasion in each region by using a classification tree analysis in the program Salford Predictive Modeler (v. 8, Minitab, State College, Pennsylvania). Classification tree analysis is a machine learning method that employs binary recursive partitioning to model categorical response variables (Breiman et al. 1984). The tree is constructed by repeatedly splitting the data into two groups (nodes) defined by a threshold value (continuous data) or category (categorical data) of a single independent variable that maximizes homogeneity of outcome (e.g., established versus not established) within the two groups created by the split (De'ath and Fabricius 2000). ...

Reference:

Predicting invasiveness of freshwater fishes imported into North America: regional differences in models and outcomes

... Bagging and Random Forests [8][9][10]. Contributions from Computer Science came later in the 1990s, with Neural Networks, Boosting and Support Vector Machine (SVM) [11,120]. ...

Reference:

Uses and opportunities for machine learning in hypertension research

... The concept of fitting an exponential curve to the tail of a sorted sample is not a new one. Breiman, Gins and Stone (1979) propose taking the maximum likelihood estimator of the tail of the sample and then fitting an exponential curve with this estimated parameter value to predict unobserved tail quantiles. Ott (1995) discusses various methods for fitting a two-parameter exponential curve to the tail of the data. ...

Reference:

Single-pass low-storage arbitrary quantile estimation for massive datasets