Mechanical Properties Prediction

Motive

To predict the mechanical properties of steel like Tensile Strength, Yield Strength, (%) Elogation & Reduction in Area (%) by its doping elements percentage.

Process

1. Used MatNavi Dataset of Mechanical Properties of Low alloyed Steels.
2. Data preprocessing like checking null values & renaming columns with appropriate names.
3. EDA by checking correlation and skewness of features. Some features are highly skewed and Correlated.

Linear Regression

4. Performing Linear Regression after removing some correlated features.
5. R² score by Linear Reg. for four dependent features are 0.82, 0.50, 0.46, 0.38.

Lasso Regression

6. As R² score is not satisfactory in this, so performed Lasso Regression considering all features becauce Lasso auto penalise the features.
7. R² score by Lasso for four features are 0.82, 0.59, 0.22, 0.19.

Random Forest Regression

8. As Lasso R² score is also unsatisfactory, hence approached for Random Forest Regression as it uses multiple Decision Trees.
9. R² score by RFG for four features are 0.92, 0.95, 0.86, 0.83.

Conclusion

The random forest regressor performs better in each category and overall as compared to Linear and Lasso Regression. Being computationally advance and highly versatile to fit itself on a complex data, this model makes for an ideal choice for prediction of mechanical properties of low-alloy steels with mean R² score of 0.89 which is significant. All the four properties of steel like Tensile Strength, Yield Strength, (%) Elogation & Reduction in Area (%) can be predicted with high accuracy.