IITM BS Machine Learning Practice Project (BSCS2008P): Scope and Tips

The Machine Learning Practice Project (BSCS2008P) is a 2 credit project course at the diploma level of the IITM BS in Data Science. You build prediction models on real data inside a Kaggle contest, and then you face a viva. It has no prerequisite, but Machine Learning Practice (BSCS2008) is a corequisite. So you take the project in the same term as MLP, or after you have passed MLP.

MLP project at a glance

The course page lists Ashish Tendulkar as the instructor. The handbook also lists the MLP project as 2 credits and makes it compulsory for the Diploma in Data Science. The two sources agree.

What the project involves

The whole project runs on Kaggle. You get a real dataset and a prediction problem, and you build a model that predicts well. The course page says the project lasts 12 to 15 weeks and is split into milestones. Each milestone covers one part of the machine learning cycle, so you move in small steps.

The course page names five parts of any ML project: data, model, cost function, optimisation and evaluation. Your work will touch all five:

• Cleaning and exploring the data with NumPy and pandas.
• Trying several models with scikit-learn.
• Choosing a loss and a metric that suit the problem.
• Tuning your models.
• Checking your results honestly before you submit.

The page does not list the milestones one by one. Your project portal shows the exact tasks and deadlines for your term.

How the project is graded

Two steps decide your grade.

1. Kaggle cutoff. The handbook says your L1 viva is scheduled only after you cross the cutoff in the Kaggle contest and declare that all your models are complete.
2. Viva. The course page says an industry expert conducts the viva. After the L1 viva, the same two cutoff system used for the other diploma projects applies. A low L1 score means you resubmit. A middle score gives a pass with a D or E grade. A high score takes you to an L2 viva for a better grade.

The full rules, including a second chance at the L2 viva, are in how IITM BS projects are graded.

Academic integrity

Treat this as a hard rule. Your notebook, your features and your viva answers must be your own work. Copied code is easy to catch in a viva when you cannot explain why a line is there. The handbook says a plagiarised project gets a U grade. You may also face disciplinary action, and you have to register again and pay the full project fee. Reading public discussions to understand an idea is fine. Pasting someone else's solution is not.

How to plan the MLP project

• Run it next to MLP. You can take both in one term. Try each week's MLP topic on your project data that same week.
• Submit a simple model early. A basic baseline on the leaderboard in the first milestone shows you how far you are from the cutoff.
• Keep a log. Write down every model you tried, its score and why you changed it. This log is your viva preparation.
• Practise explaining. Be ready to say why you picked a metric, how you handled missing values and why one model beat another.
• Do not leave it for your checkpoint term. The handbook advises finishing diploma projects in the term before your milestone term, so a failed attempt can still be redone in time. See diploma level checkpoints.

If you do not finish in the term you registered, your grade is marked I and the fee carries over to the next term. If you still do not finish in that second term, the grade becomes U and you pay again.

What to take before and after

Machine Learning Foundations and Machine Learning Techniques give you the theory. Machine Learning Practice (BSCS2008) gives you the hands-on practice you apply here. Our guide to MLF, MLT and MLP explains how the three fit together. After both diplomas, Deep Learning is the next ML course, at the BSc level.

Common questions

Can I do the MLP project before the MLP course?

No. MLP (BSCS2008) is a corequisite. You must have passed it already or take it in the same term as the project.

Does the MLP project grade matter for the BSc level?

Yes. It is one of the four diploma projects. The handbook needs a CGPA of 7.0 or more across your diploma projects before you can enter the BSc level.

What if I finish all six Data Science theory courses but not this project?

The handbook gives an Advanced Certificate in Machine Learning and Data Science to diploma students who finish all six theory courses but have one or both projects pending. It is not the full diploma, so you still need the project for that.

Is the Kaggle cutoff score printed in the handbook?

No. The handbook only says you must cross the cutoff in the contest before the L1 viva. Check the project portal for your term.