This project is interested in the problem of the classification of heavy and toxic metal elements in phosphate fertilizers according to the elements that exist in the phosphate. The idea is to use the elements and their density to train the machine learning model that classifies whether they are toxic and heavy elements and to propose models that can classify automatically. This is done by proposing three models to choose the best one. More concretely, we propose a set of learning models such as a supervised learning model KNN, a supervised learning model logistic regression, a Decision Tree supervised learning model, a Random Forest supervised learning model, an XGBoost supervised learning model, and a neural network learning model ANN. Finally, to validate the efficiency of the proposed models, they have been evaluated on a problem of classification of heavy and toxic metals in phosphate or not. The proposed solutions are not yet used, they are just proposed because they can work well with a small amount of data as in our case, and they are modifiable.