I'm a first-year BTech/B.E. computer science and data science student. I understand the basic use of machine learning models in scikit-learn, and I know enough pandas to filter datasets, but I'm missing the mathematics needed to understand how the algorithms work internally. I'm currently learning scikit-learn preprocessing and want to become more comfortable working with real-world data. My plan is to study probability, statistics, and linear algebra during September, then complete Andrew Ng's machine learning course in October. From October through November, I want to apply what I learn by cleaning, preprocessing, and modeling more complex datasets. I'm also planning to learn SQL from November through January. Does this seem like a sensible learning path, and what would experienced people change or prioritize?
2 Answers
The overall direction is solid: build the math foundation, take a structured machine learning course, and then apply it to practical datasets. Just make sure your linear algebra includes vectors, matrices, eigenvalues, and at least an introduction to matrix calculus, since those topics make the course much easier to follow. Also, don’t wait until November to begin SQL—learning it alongside pandas and preprocessing would be more efficient. SQL is a major part of working with real data and is often the fastest way to filter, join, and aggregate information before bringing it into Python.
They overlap, but they’re used in different places. SQL is usually used to query and transform data inside a database, often before you load it into Python. Pandas is better for in-memory analysis and custom transformations after extraction. In many real projects, you use both rather than choosing only one.
Your plan is reasonable, but treat it as a flexible progression rather than a strict calendar. Focus on understanding the fundamentals and completing a few end-to-end projects instead of moving on just because a month has ended. For each project, try to define the problem, inspect and clean the data, create a baseline, train a model, evaluate it with an appropriate metric, and explain the results. SQL is especially useful for fintech and data-focused roles, so learning it earlier would be worthwhile. You do not need competition-level mathematics immediately; strong school-level algebra, probability, statistics, and linear algebra are enough to start, as long as you keep deepening them over time.
I’m a little confused about the overlap. If pandas and scikit-learn can filter and preprocess data too, why is SQL still necessary?