tribhuvan0 / comparative-analysis-of-ml-algorithms

House Price Prediction using different machine learning models

Author: Tribhuvan Mishra

This project aims to predict the house price by using given set of variables and also to find which algorithm gives best performance

This repository contains the notebook associated with the task performed

• Notebook has well written code along with markdowns, where additional information is given.
• Notebook also has sufficient visualizaations.
• Instead of only using algorithms, I have used hyperparameter tuning techniques to get better results.
• A table at the last shows the comparative study of all the algorithms used.
• I have also explained the reason behind the performance of each and every model on our dataset

Steps taken

• Data visualization and exploratory data analysis
• Using different models to this regression problem
• Found the best model among the models used
• Created visualisation of performance of different models
• Also included the reasons for the prformance of different models

Machine learning models used and performance matrix

Algorithm used	MAE(Mean Absolute Error)	MSE(Mean Squared Error)	RMSE(Root Mean Squared Error)
Linear regression	5.347	43.279	6.579
Support vector machine	5.071	45.617	6.753
Random forest regressor	4.796	40.973	6.401
Random forest regressor (random search cv	3.964	25.677	5.067
Random forest regressor (grid search cv)	4.669	37.328	6.109

Future Enhancements

• We can try to work again with svm, but with normalised values of data
• It can be also deployed on web application

References

• https://trace.tennessee.edu/cgi/viewcontent.cgi?article=7047&context=utk_gradthes#:~:text=Random%20search%20can%20work%20better,better%20than%20both%20of%20them.
• https://www.researchgate.net/publication/274638232_Performance_of_Random_Forest_and_SVM_in_Face_Recognition#:~:text=The%20SVM%20achieved%20accuracy%20of,of%20SVM%2C%20RF%20and%20classification.

For running it live

• House Price prediction

Questions?

Email the author at [email protected]