The water crisis is compounded by a number of factors, including population growth. In order to assess water potability, several indicators need to be taken into account during water quality evaluation. The World Health Organisation (WHO) sets concentration standards for each parameter to ensure that it is fit for drinking. The aim of this work is to take an in-depth look at these various water parameters, which have a significant impact on human health, and to understand how they influence water quality by using advanced machine learning techniques. The methodology consists on the one hand, to build a model for predicting the potability of water and on the other hand to study the impact of certain physico-chemical factors related to human health in this potability. The study was based on the use of three machine learning algorithms, namely Decision Tree, XGBoost and Random Forest, to analyze the impact of parameters such as pH, chlorine, chlorides, turbidity, nitrates, conductivity and fluoride. The results for the prediction model are promising especially for the Random Forest algorithm which gives the best performances. Regarding the impact of physico-chemical factors in the potability, all the algorithms place pH and chlorine at the top. Other parameters such as chlorides and turbidity are also significant, although their contribution is slightly lower than that of the previous characteristics.