The K-Nearest Neighbour (kNN)-aided Predictions of Aquifer Fluoride Levels in the Rahole Area, Dadaab sub-County

Abstract

Abstract— Fluoride levels have been a challenge in the Merti aquifer, especially those that are located on the distal and peripheral fringes of the Merti. Some areas in these two zones have acceptable levels of fluorides whereas others do not have acceptable levels. Bone fluorosis has been a public health concern in the border areas and this defect has been blamed on suspected anomalous fluoride levels of some old wells. Rahole area is located in the distal Merti and is therefore a candidate for anomalous levels with a relatively high probability. Since the study area is a new settlement, the hydrogeological assessment incorporated a study into the expected levels of fluorides so that the borehole is sunk with this knowledge in mind-the expected levels of fluoride. The levels may not be suitable for humans but still be favorable for livestock. To achieve this, a Machine Learning algorithm, K-Nearest Neighbor, was used. The results generated prove over ninety percent accuracy levels of predictions. The study area is found to be free of fluoride levels deemed harmful to human life.