PythonWeek-GraduateAdmission

Students all across the world go through a phase where they have to make a decision whether they wish to study further in order to gain in-dept knowledge about the stuff in their field or whether they start working. Well that depends on an individual and what he/she wishes to do but if you are someone who wishes to pursue further studies, this project might help you out.

What is the Project about?

It is a web-app that makes use of Machine learning inorder to determine whether a student will be able to get into a Graduate Admission based on metrics such as their GRE scores, TOEFL scores, their CGPA and other relevant stuff.

What did we use to build it?

The primary language for modeling and data analysis was Python. We made use of Sci-Kit learn library in order to make our model which is a linear regression model. We made use of Libraries like Matplotlib and Seaborn for visualizing our data.

About the Data-set

I used a data set off of Kaggle. It contained data of real students and their GRE scores, TOEFL scores and other relevant data which is important during Admission in any Graduate Program.

How was the project built?

1. I started off by getting the data from Kaggle and used the Kaggle notebooks to create the model. After importing the data i looked at the different columns in our data.
   
2. I then visualized the various columns of our data using various graphs in order to determine the columns that i wished to use to make our model.
   
3. Once i had my columns, i split the data into training and testing set using the train-test split and made my Linear Regression model using the Sci-Kit learn Library.
   
4. Once the model was ready, i put it in a pickle file.
   
5. I then shifted to Colab Notebook where i deployed the model on web using PyNgrok and Streamlit. The front end of the Web app was all made using Stream lit and i then used Pyngrok to host it on web.
   
6. In the end i tested the web app with the inputs from the test set that we created and the model gave results very close to the results we had in our original data.
   

Image Gallery

Conclusion

The model was able to make some very good predictions. The model accuracy was somewhere around 94-95% which is quite decent and even on the data which the model had never seen before, it was able to make some good predictions.