Lattice Thermal Conductivity calculations of Inorganic compounds are highly resource intensive and time consuming DFT calculation. Hence, discovery of new novel inorganic compounds is hindered because of it. Machine Learning approach can help in surpassing need of such DFT calculations and speed up the material search. For this purpose, Random Forest model is developed to predict Lattice Thermal Conductivity with Temperature.
In this project, Machine Learning Model (Random Forest) is build to predict Lattice Thermal Conductivity of Inorganic Compounds at various temperatures. Workflow of Project is as follows:
- Preparing Feature Dataset from Initial Dataset : Initial Dataset contains information about Lattice Thermal Conductivity vs Temperature values for each Compound. There are total 119 Compounds present in Dataset.
- Generating Crystal and Compound features of Compounds : For this step, another Java library MAGPIE (Materials Agonnistic Platform for Informatics and Exploration) is used.
- Exploratory Data analysis (EDA) of features : All the features obtained are analysed and only important features are selected to train Machine Learning models. EDA is required to analyse the presence of bias in data.
- Training and Testing Machine Learning Models : Following Machine Learning Model are trained and validated for LTC prediction task.
Code of Project and Dataset is taken from the following reference article:
Paper : Lattice Thermal Conductivity: An Accelerated discovery guided by Machine Learning Code : GitHub Repo link