Machine learning models and dataset for the prediction of Cr6+ removal of aqueous solutions using the pine cone residue

Description

The experimental results of the adsorption of Cr6+ from aqueous solutions using the pine cone residue and the machine learning models used to investigate the experimental parameters are available on this page. Three adsorption conditions are optimized: contact time, pH, and initial solution concentration. The file “Pinha.csv” presents the results of the experiments considering various adsorption conditions. In addition, three machine learning models are employed to predict the behaviour of the experimental results: a multiple linear regression model, a decision tree model, and a random forest model. The machine learning models are in the “Machine learning models.ipynb” file. The maximum Cr6+ removal obtained with the pine cone residue is near 91%, and the machine learning models presented a high correlation coefficient of over 0.9, highlighting the potential of this type of methodology to enhance experimental studies.

Steps to reproduce

To reproduce this research, scientists should open the Python code (“Machine learning models.ipynb”) with an IDE such as Jupyter Notebooks or VS Code. The Pandas, Sklearn, Matplotlib, and Plotly libraries are required for the code to work. The code is divided into subsections: “1) Import libraries”, “2) Import and pre-process the data”, “3) Multiple Linear Regression”, “4) Decision Tree”, “5) Random Forest Model”, “6) Plotting the results”, “7) Plotting the decision tree”. Sections 1 and 2 are standard for all the machine learning models, available in sections 3, 4, and 5. The visualization of the results is presented in sections 6 and 7. Notice that the code automatically reads the results on the database file (“Pinha.csv”); hence, both the code and the csv file should be placed in the same folder. If the user cannot access the ipynb program, a “.py” file that can be opened on traditional Python IDEs is also provided.

Institutions

Categories

Funding

Licence