A Hybrid Probabilistic Ensemble based Extreme Gradient Boosting Approach For Breast Cancer Diagnosis

Exploring the dominant features and data-driven detection of polycystic ovary syndrome through modified stacking ensemble machine learning technique

“…Furthermore, using feature selection strategies, the majority of previous studies randomly picked a specified number of features. For example, Bharti et al [65] applied ML classifiers with ten statistically significant features based on p-values, Inan et al [37] proposed to use most significant top twelve features, Danaei et al [29] had acquired best accuracy employing 28 features selected using Random Forest embedded feature selection technique and so on. However, hardly any study has investigated at how changing the numbers and combinations of features selected using that same feature selection method can affect the prediction result.…”

“… Chandrasekar et al [84] , Bahad et al [85] , Deif et al [86] eXtreme Gradient (XG) Boosting This approach is scalable and efficient form of gradient boosting that improves on two fronts: tree construction speed and a novel distributed algorithm for tree searches [87] Heart disease detection, chronic kidney disease diagnosis, breast cancer detection etc. Ashish et al [88] , Ogunleye et al [89] , Inan et al [90] Adaptive Boosting classifier It's an adaptive classifier that leverages the results of various weak learning algorithms to substantially enhance performance and provide an effective predictor for the boosted classifier's final output [91] Endometrial cancer prediction, Hepatitis disease detection, cancer classification etc. Wang et al [92] , Akbar et al [93] , Lu et al [94] Categorical Gradient (CAT) Boosting It is an implementation of Gradient Boost classifier that employs ordered boosting with categorical features and uses binary decision trees as underlying predictors [95] Parkinson's disease prediction, COVID-19 detection from blood samples, diabetes risk prediction etc.…”

“…Since PCOS is associated with a wide range of symptoms as features, a few studies have emphasized on employing various feature reduction approaches before using machine learning models to accelerate the training process. For example, Inan et al [37] suggested a strong sampling technique that includes both oversampling and undersampling procedures to boost minority samples; then applied two types of feature selection techniques: Chi- Square test for categorical and Analysis of Variance (ANOVA) test for numerical attribute selection; and then applied six types of machine learning classifiers where XGBoost classification model outperformed others. In another relevant article in this domain, Nandipati et al [38] used RFE-LR, RFI-ECT, SelectKBest/Chi2 and Forward Backward propagation techniques to find the top 10 and 24 features from all 42 features in the dataset, and then applied seven types of traditional ML classifiers in two different types of implementation platforms: Python-Scikit Learn package and RapidMiner; in addition, performance comparisons between various classifiers were assessed utilizing complete (40 features) and selected features (10 and 24 features) to find the best performing classifier.…”

A systematic review and future research agenda on detection of polycystic ovary syndrome (PCOS) with computer-aided techniques

“… Feature selection techniques Ref Sequential backward selection (SBS) [37] Pearson method [37] Random Forest embedded feat. selection [37] Recursive Feature Elimination (RFE) [38] , [54] Chi-Square feature selection [38] , [56] Forward Selection & Backward Elimination Technique [38] Genetic Algorithm [51] Decision Tree [52] Filter Method [53] Univariate Feature Selection Method [54] Statistical Analysis (t-test) [55] Analysis of variance (ANOVA) Test [56] , [59] Neural Fuzzy Rough Set (NFRS) & Artificial Neural Network (ANN) [57] Correlation based Feature Selection (CFS) [57] Principal Component Analysis (PCA) [57] , [58]

Figure 11 Ven Diagram of applied research methodologies.

…”

Explainable extreme boosting model for breast cancer diagnosis

“…Inan et al [7], evaluated the synthetic minority oversampling technique for optimization of the XGBoost model for breast cancer diagnosis. The study developed the XGBoost with a dataset collected from the UCI repository.…”