English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Tools annual

Global: Zendy Free Plan

Global: Zendy Tools monthly

Global: Zendy Plus monthly

Most real-world classification problems deal with imbalanced datasets, posinga challenge for Artificial Intelligence (AI), i.e., machine learningalgorithms, because the minority class, which is of extreme interest, oftenproves difficult to be detected. This paper empirically evaluates tree boostingmethods' performance given different dataset sizes and class distributions,from perfectly balanced to highly imbalanced. For tabular data, tree-basedmethods such as XGBoost, stand out in several benchmarks due to detectionperformance and speed. Therefore, XGBoost and Imbalance-XGBoost are evaluated.After introducing the motivation to address risk assessment with machinelearning, the paper reviews evaluation metrics for detection systems or binaryclassifiers. It proposes a method for data preparation followed by treeboosting methods including hyper-parameter optimization. The method isevaluated on private datasets of 1 thousand (K), 10K and 100K samples ondistributions with 50, 45, 25, and 5 percent positive samples. As expected, thedeveloped method increases its recognition performance as more data is givenfor training and the F1 score decreases as the data distribution becomes moreimbalanced, but it is still significantly superior to the baseline ofprecision-recall determined by the ratio of positives divided by positives andnegatives. Sampling to balance the training set does not provide consistentimprovement and deteriorates detection. In contrast, classifier hyper-parameteroptimization improves recognition, but should be applied carefully depending ondata volume and distribution. Finally, the developed method is robust to datavariation over time up to some point. Retraining can be used when performancestarts deteriorating.

Tree Boosting Methods for Balanced andImbalanced Classification and their Robustness Over Time in Risk Assessment