K Nearest Neighbors with Adaptive Weighting to Forecast Long-Term Outcomes of Stem Cell Therapy in Regenerative Medicine

In recent years, stem cells are the fundamental building blocks of life which possess immense potential in regenerative medicine by offering promising prospects for Africa's tissue regeneration with stem cell therapy. Traditional approaches for forecasting stem therapy in regenerative medicine has faced several challenges which include lack of interpretability, model drift and overfitting. Therefore, this research proposes K Nearest Neighbors (KNN) with Adaptive Weighting (AW) namely KNNAW for forecasting stem therapy in regenerative medicine which is employed on Stem Cell Therapy Outcome Database (SCTOD) a preprocessed by using standard scalar normalization. Then, these data are used for selecting relevant features by using Gaussian mixture model and mutual information which handles complex distributions. After that, evaluation is done by using proposed KNNAW where KNN is utilized for finding nearest neighbor to new patient based on selected features and AW is used to assign weights to each neighbor based on similarity to new patient. Finally, prediction is done by using Support Vector Regression (SVR) which effectively captures nonlinear relationships between variables. The proposed KNN achieved better accuracy (0.921), specificity (0.764), sensitivity (0.992) and F1-score (0.884) when compared with existing DenseNet121.