Hybrid(Penalized Regression and MLP) Models for Outcome Prediction in HDLSS Health Data

A new machine learning model has been developed for early diabetes prediction, utilizing the BRFSS dataset, which includes over 253,680 records. The model employs various supervised learning techniques, including ensemble methods like stacking, achieving a strong ROC-AUC performance of approximately 0.96 with models such as Random Forest, XGBoost, CatBoost, and LightGBM.

A new classification framework named Smart Bayes has been introduced, which integrates likelihood-ratio-based generative features into a logistic-regression-style discriminative classifier. This approach allows for data-driven coefficients on density-ratio features, enhancing class separation compared to traditional methods like Naive Bayes and logistic regression.

A new method called Decision Tree Embedding (DTE) has been proposed, which utilizes the leaf partitions of a trained classification tree to create an interpretable feature representation. This approach aims to reduce the high estimation variance typically associated with single decision trees while maintaining interpretability and efficiency in classification tasks.

A new deep learning architecture named Rep3Net has been proposed to enhance molecular bioactivity prediction in early-stage drug discovery. This model integrates traditional molecular descriptor data with spatial and relational information through graph-based representations and contextual embeddings generated by ChemBERTa from SMILES strings. The model has shown reliable predictions on the Poly [ADP-ribose] polymerase 1 (PARP-1) dataset, which is vital for DNA damage repair in cancer therapies.

A new machine learning framework has been developed to optimize stroke risk prediction, utilizing ensemble modeling and explainable AI techniques. This framework achieved an impressive accuracy of 99.09% on the Stroke Prediction Dataset through a comprehensive evaluation of various models and data preprocessing methods, including Random Over-Sampling to address class imbalance.

CLAPS has been introduced as a novel posterior-aware conformal regression method that utilizes a Last-Layer Laplace Approximation combined with split-conformal calibration, resulting in narrower prediction intervals while maintaining target coverage, particularly beneficial for small to medium datasets.