Data Science

A collection of data analysis, forecasting and visualization notebooks that apply Python's data-science stack (pandas, NumPy, scikit-learn, TensorFlow/Keras, NLTK, matplotlib) to practical questions.

Credentials must be supplied by the user: notebooks that call authenticated APIs read their keys from environment variables. See the .env.example file in each folder. No keys are stored in this repo.

Structure

notebooks/

ml-fundamentals/ is a quick-reference section: compact reference implementations of standard techniques on toy or built-in datasets, not original research.

Folder	Notebooks	What they cover
ml-fundamentals/	linear & logistic regression, decision tree, random forest, gradient boosting, k-NN, SVM, PCA (plus pca-from-scratch, which builds PCA from eigen-decomposition and checks it against scikit-learn), t-test, pandas basics, loops	Compact reference implementations of core techniques, mostly on the Iris dataset or synthetic data
signal-processing/	fourier-transform, lstm-sequence-windowing	FFT spectrum analysis; preparing time-series windows for recurrent models
nlp-vision/	nltk-tokenization, opencv-image-basics	Text pre-processing; image loading and display in headless Jupyter
nlp-social/	twitter-timeline-word-frequency, reddit-wallstreetbets-word-frequency	Topic discovery from social-media text via token frequencies. 🔑 API credentials required
data-ingestion/	pdf-table-extraction	Pulling tables out of PDF reports with tabula
finance-capitalstream/	correlation matrix, ETF ratios, summary stats, trend regression, affinity-propagation market structure, LSTM-autoencoder anomaly detection, LSTM forecasting, crypto market data, crypto lending yield, FRED macro indicators	CapitalStream Insight, a personal investment-analytics project on public market data. 🔑 The FRED notebook needs a Nasdaq Data Link key

Each notebook opens with a short header covering the problem, approach and data, plus attribution where it adapts a published tutorial. The finance notebooks get their risk/return metrics from the local scripts/risk_metrics.py module, so they run without any external course code.

learning-with-kids/

Seven beginner Python programs for a child aged 6–8 with a parent, run in Thonny. Parent guides are in French; code and comments are in English. They cover input/output, functions, loops, if/else, variables and lists, turtle graphics, and a guess-the-number game. test_lessons.py runs every program with scripted answers, and the turtle drawings under a virtual display.

datasets/

Sample or public data only, never client or proprietary data:

• synthetic/crypto-lending/: generated lending and deposit histories used by crypto-lending-yield.ipynb. The values and addresses are fake.
• sample-images/: a generated test image for the OpenCV notebook.
• sample-pdfs/: a synthetic one-page report (sample_table.pdf) and the script that generates it, used by the PDF table-extraction notebook.

scripts/

Path	What it is
risk_metrics.py (+ test_risk_metrics.py)	Original, dependency-light module: annualised return and volatility, Sharpe ratio, max drawdown and summary_stats(). Used by the finance notebooks and unit-tested with pytest.
cheminformatics/smiles_to_molblock.py	SMILES → MOL block with 2-D coordinates (RDKit)
ml-from-scratch/svm_from_scratch.py	A linear SVM written from scratch in NumPy. Adapted from sentdex, "Practical Machine Learning with Python"
ml-from-scratch/stock_price_regression.py	Linear vs SVR price forecasting on engineered features. Adapted from sentdex, "Practical Machine Learning with Python"
ml-from-scratch/sklearn_digits_quickstart.py	The scikit-learn estimator API on handwritten digits. Adapted from the scikit-learn "Introduction to machine learning" tutorial
matplotlib-basics/matplotlib_chart_types.py	Line, bar, histogram, scatter and plot-from-file reference. Adapted from sentdex, "Matplotlib tutorial series"
matplotlib-basics/live_graph.py + signal_generator.py	A live-updating chart fed by a synthetic signal generator. live_graph.py adapted from sentdex, "Matplotlib tutorial series"
matplotlib-basics/basemap_world_map.py	Minimal world map. Adapted from sentdex's Basemap tutorial
finance-2016/*.py	Early price-download, charting and CSV-export scripts, ported from retired Yahoo/pandas APIs to yfinance. Adapted from sentdex's Matplotlib and Pandas tutorial series (see each file's header)

Every tutorial-derived script names its source in an "Adapted from:" line in its header.

requirements.txt

Python dependencies.

Courses completed

Structured courses behind the techniques used in these repos. Only the course links are listed here: the course materials belong to their authors and LinkedIn, and aren't redistributed.

Course (LinkedIn Learning)	Instructor	What it covered
Machine Learning with Python: Foundations	Frederick Nwanganga	The ML workflow end to end: collecting, exploring and preparing data, then building, evaluating and interpreting models in Python
Machine Learning with Scikit-Learn	Michael Galarnyk	Supervised learning (linear/logistic regression, decision trees, bagged trees, random forests), unsupervised learning (k-means, PCA) and scikit-learn pipelines
Python for Engineers and Scientists (exercise repo)	Michele Vallisneri	Fast numerics with NumPy, SciPy, Numba and Cython; correctness with SymPy, ODE solvers and interpolation; JSON/pandas/HDF5 data handling, scripting and Snakemake workflows
Building and Deploying Deep Learning Applications with TensorFlow	Adam Geitgey	Building, training, checkpointing, visualising (TensorBoard) and deploying a TensorFlow model to a cloud prediction service
Using Python for Automation	— (2022 edition)	File and directory automation, web scraping, and browser automation with Selenium

Philosophy

Each notebook aims to show not just the technique but the reasoning behind it: what question is being answered, why this method, and what the result means in practice.

Status

🚧 Under active development.