Learn data science from scratch

{{whyLabel}}: Linear algebra is the language of data; matrices and vectors are how computers store and process datasets.

{{howLabel}}:

• Learn vector operations (addition, dot product) and matrix multiplication.
• Understand Transpose, Inverse, and Determinants.
• Study Eigenvalues and Eigenvectors as they are critical for Principal Component Analysis (PCA).

{{doneWhenLabel}}: You can manually multiply a 2x2 matrix and explain what a vector represents in a coordinate system.

{{whyLabel}}: Calculus, specifically derivatives, is the engine behind 'Gradient Descent'—the method models use to learn from errors.

{{howLabel}}:

• Focus on Derivatives and the Power Rule.
• Understand the Chain Rule for nested functions.
• Learn Partial Derivatives to handle functions with multiple variables (features).

{{doneWhenLabel}}: You can explain how a derivative helps find the minimum of a cost function.

{{whyLabel}}: You must be able to summarize large datasets into meaningful numbers like averages and spreads.

{{howLabel}}:

• Practice calculating Mean, Median, and Mode.
• Understand Variance and Standard Deviation to measure data 'noise'.
• Learn about Percentiles and Quartiles for identifying outliers.

{{doneWhenLabel}}: You can describe a dataset's distribution using its five-number summary (Min, Q1, Median, Q3, Max).

{{whyLabel}}: Data science is about predicting the likelihood of events; Bayes' Theorem is the foundation of many classification algorithms.

{{howLabel}}:

• Study Conditional Probability (P(A|B)).
• Solve problems using Bayes' Theorem (e.g., medical test accuracy).
• Learn about Probability Distributions (Normal, Binomial, Poisson).

{{doneWhenLabel}}: You can calculate the probability of an event given new evidence using the Bayesian formula.

{{whyLabel}}: A standardized environment ensures your code runs reliably and manages all your data science libraries.

{{howLabel}}:

• Download and install the 'Anaconda Distribution' (Individual Edition).
• Install 'Visual Studio Code' and the 'Python' extension.
• Create your first virtual environment using the command: conda create -n ds_env python=3.11.

{{doneWhenLabel}}: You have a working Python environment and can launch a Jupyter Notebook from VS Code.

{{whyLabel}}: You need to automate tasks and handle data structures before using advanced libraries.

{{howLabel}}:

• Practice Lists, Dictionaries, and Sets.
• Write 'For' loops and 'If-Else' logic.
• Learn List Comprehensions for concise data processing.
• Define Functions with arguments and return values.

{{doneWhenLabel}}: You can write a script that processes a list of numbers to return only the even ones squared.

{{whyLabel}}: Most company data lives in databases; SQL is the only way to retrieve it.

{{howLabel}}:

• Learn SELECT, FROM, and WHERE clauses.
• Master JOINs (Inner, Left, Right) to combine tables.
• Use GROUP BY and Aggregate functions (SUM, AVG, COUNT).
• Practice on 'SQLZoo' or 'Mode Analytics' free tutorials.

{{doneWhenLabel}}: You can write a query to find the top 5 customers by total spend from two related tables.

{{whyLabel}}: Pandas is the industry standard for tabular data; NumPy handles the underlying numerical math.

{{howLabel}}:

• Learn to load CSV files into a 'DataFrame'.
• Practice filtering, sorting, and grouping data.
• Handle missing values using .fillna() or .dropna().
• Use NumPy arrays for fast mathematical operations on columns.

{{doneWhenLabel}}: You can load a dataset and calculate the average value of a column grouped by a specific category.

{{whyLabel}}: Visuals reveal patterns (like correlations) that raw numbers cannot show.

{{howLabel}}:

• Build Line charts for trends and Bar charts for comparisons.
• Use Histograms to see the distribution of a single variable.
• Create Scatter plots to find relationships between two variables.
• Use Seaborn's heatmap() to visualize correlation matrices.

{{doneWhenLabel}}: You have a notebook showing a correlation between two variables in a sample dataset.

{{whyLabel}}: EDA is the 'detective work' where you find errors, outliers, and interesting trends before modeling.

{{howLabel}}:

• Check for data types and null values using .info() and .describe().
• Identify outliers using Box plots.
• Check for class imbalance (e.g., are 99% of transactions 'not fraud'?).
• Formulate 3 hypotheses about the data and test them with visuals.

{{doneWhenLabel}}: You have a completed EDA report on a dataset (e.g., Titanic or House Prices).

{{whyLabel}}: Regression is the foundation of predictive modeling, used for forecasting continuous values like prices.

{{howLabel}}:

• Split data into 'Training' and 'Testing' sets using train_test_split.
• Train a model using Scikit-Learn's LinearRegression().
• Evaluate using Mean Squared Error (MSE) and R-squared.

{{doneWhenLabel}}: You have a model that predicts a numerical value with an R-squared score.

{{whyLabel}}: Classification is used for binary outcomes (Yes/No, Spam/Not Spam).

{{howLabel}}:

• Understand the Sigmoid function.
• Train a LogisticRegression model.
• Create a Confusion Matrix to see True Positives vs False Positives.

{{doneWhenLabel}}: You can explain the difference between Precision and Recall for your model.

{{whyLabel}}: Tree-based models are highly powerful and handle non-linear data better than simple regression.

{{howLabel}}:

• Learn how a Decision Tree splits data based on 'Entropy' or 'Gini Impurity'.
• Use 'Random Forest' (an ensemble of trees) to improve accuracy and reduce overfitting.
• Visualize a single tree using plot_tree.

{{doneWhenLabel}}: You have trained a Random Forest and identified the 'Feature Importance' (which variables matter most).

{{whyLabel}}: Sometimes you have data but no labels; clustering finds hidden groups (e.g., customer segments).

{{howLabel}}:

• Use the K-Means algorithm.
• Find the optimal number of clusters using the 'Elbow Method'.
• Visualize the clusters in a scatter plot.

{{doneWhenLabel}}: You have grouped a dataset into 3-5 distinct clusters based on similarity.

{{whyLabel}}: Real-world data is messy; finishing a project proves you can handle the full pipeline from cleaning to prediction.

{{howLabel}}:

• Choose a dataset (e.g., 'Telco Customer Churn' or 'NYC Taxi Trips').
• Perform cleaning, EDA, Feature Engineering, and Modeling.
• Document every step in a clean Jupyter Notebook.

{{doneWhenLabel}}: You have a final notebook that achieves a specific accuracy or error metric.

{{whyLabel}}: GitHub is your 'technical resume' where recruiters verify your coding ability.

{{howLabel}}:

• Create a GitHub account.
• Learn basic Git commands: git init, add, commit, push.
• Upload your best 2-3 projects with a professional README.md explaining the problem, solution, and results.

{{doneWhenLabel}}: Your GitHub profile has at least two repositories with clear documentation.

{{whyLabel}}: You need to translate your technical projects into business value to get interviews.

{{howLabel}}:

• Use the 'X-Y-Z' formula: 'Accomplished X as measured by Y, by doing Z'.
• Highlight specific tools (Python, SQL, Scikit-Learn).
• Include a link to your GitHub and LinkedIn.

{{doneWhenLabel}}: You have a one-page PDF resume tailored for Junior Data Scientist roles.