This repository contains all the R codes written specifically for contrusting machine learning models to predict the initial dischagre capacity (IC) and 20th cycle end discharge capacities of 102 doped spinel cathode materials. This is given to aid the dicussion in the manuscript " Insight Gained from Using Machine Learning Techniques to Predict the Discharge Capacities of Doped Spinel Cathode Materials for Lithium-ion Batteries Applications"
In this repository, six different machine learning techniques are written including two linear models and 4 non-linear models:
Linear models:
Penalised_regression_codes.R (for Ridge and Lasso penalised regression)
Non-linear models:
- Artificial Neural Network.R
- SVM_codes.R (Support Vector Machine)
- Random_forest.R
- Decision_tree.R
- Gradient_Boosting_Machine.R
Dataset:
LMO.csv
Operating system: windows 10, 64bit Software: R version 3.6.0
For the required R libraries, please see each the first line discussion in each code file
The Codes for optimising the hyperparameters are also included in the files as the reference to the labelled optimised values.
Please go ahead to the training and testing section for quick access of the models' results
1st column is the ratio of dopant in the material formula (M)
2nd column is the ratio of managanese in the material formula (Mn)
3nd column is the electronegativity of the dopant element (M_EN)
4th column is the molar mass of the material (Mr)
5th column is the lattice constant a of the material's crystal structure, obtained from the reported X-ray diffraction spectrums (LC_a)
6th column is the current density applied for both charging and discharging the battery (CD)
7th column is the initial discharge capacities (IC)
8th column is the 20th cycle end discharge capacities (EC)
Root_mean_Square_error for the test-set regression prediction for both IC and EC