Predicting composition of Cr-Mo alloy

Problem description

Using the regression models in a real thermodynamic system.

Applied two supervised algorithms, Linear regression and K-Nearest Neighbors to predict the composition values at various eta values.

---

What does eta values mean?

These are dimensionless number, that represent normalized energy parameters which corresponds to:

• eta1 - first neighbor cluster
• eta2 - second neighbor cluster
• eta3 - third neighbor cluster
• eta4 - fourth neighbor cluster

---

What are the other contents in the dataset?

The dataset contains: u0, u1, u2, u3, u4, eta1, eta2, eta3, eta4. Here

• u0 - Composition
• u1 - Correlation function for I-neighbor pair
• u2 - Correlation function for II-neighbor pair
• u3 - Correlation function for triangle
• u4 - Correlation function for tetrahedron cluster

---

Results

• Accuracy of linear model on training dataset : 49.709004178502795 %
• Accuracy of linear model on testing dataset : 51.05961513782863 %
• Accuracy of K-Nearest Neighbor model on training dataset : 59.11328125000001 %
• Accuracy of K-Nearest Neighbor model on testing dataset : 51.05961513782863 %

Using the efficient model, preferbly the KNN model can be used to predict the composition values at different eta values.