Regression with feature engineering
Deliverable: A price or demand prediction model with documented feature engineering, a baseline comparison, cross-validation results and residual analysis identifying where the model fails.
Course
Machine Learning
Classical machine learning done properly — feature engineering, honest evaluation, and the discipline to notice when a good score is an artefact rather than a result.
- Duration: 4 months
- Classroom · Online live
- Level: intermediate
What you will be able to do
- Frame a business problem as a supervised, unsupervised or non-machine-learning task
- Engineer features that carry signal without leaking information from the future
- Train and tune regression and classification models on real, messy data
- Choose an evaluation metric that reflects the cost of each error type
- Detect data leakage, overfitting and spurious correlation in your own work
- Explain a model's prediction and state where it is likely to fail
- Package a trained model behind a simple interface
Who this course is for
- Graduates in engineering, science, mathematics, statistics or economics
- Working analysts moving from reporting into modelling
- Comfort with Python and basic statistics is assumed rather than taught from scratch
- Class 12 mathematics is sufficient for the theory covered
Syllabus
8 modules · 4 months
Module 1. Framing and baselines
- Deciding whether a problem needs machine learning at all
- Supervised, unsupervised and reinforcement framing
- Building a baseline before building a model
- Train, validation and test splits, and respecting time order
- How data leakage happens, with worked examples
Module 2. Data preparation
- Profiling a dataset before modelling it
- Missing data strategies and their consequences
- Encoding categorical variables
- Scaling, and which models care about it
- Handling class imbalance honestly
Module 3. Regression
- Linear regression from first principles
- Regularisation with ridge and lasso
- Residual analysis and what it reveals
- Error metrics and choosing between them
Module 4. Classification
- Logistic regression and reading its coefficients
- Decision trees and their instability
- Random forests and gradient boosting
- Precision, recall, F1 and ROC-AUC, and when each misleads
- Threshold selection against the cost of a false positive
Module 5. Feature engineering
- Domain-driven feature creation
- Interaction and polynomial features
- Time-based features without leaking the future
- Feature selection and importance measures
- Pipelines that apply the same transforms at inference
Module 6. Model selection and tuning
- Cross-validation strategies including time-series splits
- Grid, random and Bayesian search
- The bias-variance trade-off, demonstrated
- Learning curves and diagnosing underfitting or overfitting
Module 7. Unsupervised methods
- k-means, hierarchical clustering and DBSCAN
- Choosing the number of clusters, and whether clusters are real
- Dimensionality reduction with PCA
- Anomaly detection
Module 8. Explainability and deployment basics
- Feature importance, partial dependence and SHAP
- Explaining a prediction to a non-technical stakeholder
- Serialising a model and serving it behind an API
- Monitoring for drift after deployment
Tools and technologies you will use
- Python
- NumPy and pandas
- scikit-learn
- XGBoost
- matplotlib and seaborn
- SHAP
- Jupyter
- Git
Projects you will build
Classification with imbalanced data
Deliverable: A classifier on genuinely imbalanced data, with a justified metric choice, threshold selection tied to the cost of each error type, and an honest confusion matrix discussion.
Leakage hunt
Deliverable: A supplied dataset containing deliberate leakage that you identify and remove, with before-and-after scores and a written explanation of how the leak inflated the original result.
End-to-end model with explanation
Deliverable: A trained model packaged behind a simple interface, with SHAP explanations for individual predictions and a one-page summary written for a non-technical reader.
Where this course can take you
- Machine Learning Engineer
- Data Scientist
- Applied ML Developer
- Quantitative Analyst
- Senior Data Analyst
Duration, modes and fees
- Duration
- 4 months
- Delivery modes
- Classroom · Online live
- Fees
- Share your details for the current fee
- Fees vary by batch and delivery mode. Share your details and an advisor will confirm the current fee.
Placement assistance
Every student gets placement assistance — that is what 100% placement assistance means. It is support for all, not a job for all. We do not promise a specific salary, a specific number of interviews, or placement at any named company, and you should be wary of anyone who does.
What is included
- A place in the monthly placement drive, held every third Saturday
- The readiness programme every second Saturday — mock interviews and preparation
- CV review against the specific roles you are targeting
- Portfolio review, so your project work is presented the way a reviewer will read it
- Access to the vacancy pool employers send directly to the Samyak network
- Guidance on which roles realistically fit your background and which do not
- A place in the next drive, with coaching, if you are not selected in this one
What is not included
- Any guarantee of a job, an interview, or a particular salary
- Placement at a named or partner company
- Applying to jobs on your behalf
- Support before you have completed the course and its project work
- Visa, relocation or overseas placement assistance
Baselines first, always
Every modelling project in this course requires a baseline before a model. Predict the mean. Predict last month. Predict the majority class.
You are not permitted to report a model result without one.
This is unglamorous and it develops judgement faster than anything else in the syllabus. A gradient boosting model that barely beats “predict last month” is not a success, and noticing that early is what separates useful machine learning from expensive theatre.
The leakage project
One of the four projects hands you a dataset with deliberate leakage in it and asks you to find it.
This is the most disliked and most useful exercise we run. Data leakage — a feature that encodes the answer, a split that lets the future inform the past — is the single most common reason a model looks excellent in development and collapses in production.
It is also almost never taught explicitly. Most learners meet it for the first time in a job, having reported a result they cannot reproduce. Meeting it in a classroom, where finding it is the assignment, is considerably cheaper.
Tabular data, deliberately
This course stays with structured tabular data rather than branching into images and text.
That is a deliberate scope decision, not a limitation. The large majority of business machine learning is tabular — churn, pricing, demand, risk, fraud — and the skills that matter there are feature engineering, honest evaluation and communicating uncertainty. Those transfer to deep learning; the reverse is less reliably true.
If your interest is specifically in images, text or generative systems, our AI course covers that ground properly.
What we will not claim
That four months makes you a machine learning engineer at a product company. It does not. Those roles typically want software engineering depth and production experience alongside modelling.
What it does is make you genuinely competent at building and evaluating models on real data, with a portfolio that demonstrates judgement rather than just accuracy. For an analyst moving toward modelling work, that is the step that matters.
Questions
Machine Learning — frequently asked questions
How is this different from your data science course?
The data science course is seven months and covers statistics, SQL, exploratory analysis, time series and communication alongside modelling. This is a focused four-month track on the modelling itself, assuming you already have Python and some statistics. Choose data science if you are starting broader; choose this if modelling is the specific gap.
How is this different from your AI course?
The AI course covers classical machine learning and then goes further into deep learning, transformers and generative systems across six months. This stays with classical machine learning on tabular data, which is what the large majority of real business modelling work actually involves.
Do I need to be strong at mathematics?
Class 12 mathematics is enough for what is covered. The theory is taught geometrically and practically rather than through proofs. What matters far more than mathematical background is scepticism — the instinct to ask whether a result could be an artefact — and that is built through practice rather than prerequisites.
Why does the course spend a whole project on data leakage?
Because it is the most common cause of models that look excellent in development and fail in production, and it is almost never taught explicitly. A 96 percent accuracy score that comes from a feature containing the answer is worse than a 70 percent score you can trust. Learning to spot it is a genuine differentiator in interviews.
Enquire about Machine Learning
Three details is all we need. A course advisor will call you back.
Related courses
- Data Science
Statistics, Python, SQL and machine learning taught as one connected discipline, with the emphasis on framing a problem correctly and knowing when a result is not real.
- Artificial Intelligence
Python, machine learning foundations and applied generative AI in one track — ending with a retrieval-augmented application you have built, evaluated and can explain end to end.
- Python Programming
Python taught as a working tool rather than a syntax tour — you finish able to read unfamiliar code, automate real tasks, call APIs and write tests that catch your own mistakes.
Next step
Talk to a course advisor
Tell us what you want to learn and we will help you pick the right course, batch and mode.