Crypto Data Science and ML with Python

Explore blockchain fundamentals with Solidity and Ethereum, learn how decentralized applications and smart contracts enable trustless value transfers, immutable ledgers, and consensus across nodes to prevent double spending.

Explore how bitcoin mining uses decentralized computation to confirm transactions, issue new bitcoins, and secure blocks through hash, nonce, proof of work, and mining difficulty.

Explore supervised learning, where a model learns a function mapping input x to output y from labeled training data, then predicts y for new x.

Preview a beginner-friendly project to predict cryptocurrency prices by building linear and polynomial regression models from blockchain features like transactions, market cap, and hash rate using API CSV data.

Discover linear regression as a simple supervised learning method that predicts a continuous value from a linear relationship between features and target, optimized by least squares.

Collect blockchain data from blockchain.com charts, export csv features like market price, transactions, hash rate, and difficulty, then prepare to join datasets with pandas for machine learning.

Remove the invalid first row from the pandas data frame, using range indexing and drop operations to clean the data, prep for normalization and scaling, and set up future visualization.

Visualize data by plotting time on the x-axis and multiple features on the y-axis, after ensuring correct data types with pandas and converting time to datetime for accurate charts.

Learn to define x and y for a bitcoin price prediction by selecting the number of transactions total as the feature and market price as the target, with proper indexing.

Build your first machine learning model with linear regression, fit it on your data, and visualize the best-fit line with matplotlib, then explore nonlinear, polynomial regression.

Build polynomial regression by transforming x with polynomial features of a chosen degree and fitting a linear model on the transformed data, then test degrees to find the optimal fit.

Preview a clustering project that uses unsupervised learning to group cryptocurrencies by features from the CryptoCompare API, with data cleaning, PCA, elbow curve to determine optimal K, and 3d visualization.

Explore unsupervised learning as a key branch of machine learning, with applications in recommendation systems, grouping customers by attributes, and pattern discovery, including principal component analysis and cluster analysis.

Cluster cryptocurrencies using unsupervised learning in Python with the CryptoCompare API in Google Colab. Convert crypto features into a Pandas DataFrame and CSV for easy analysis.

Clean and prune the crypto data frame by dropping noisy columns, keep meaningful features like algorithm and proof type, and prepare text features for clustering-ready data in Python.

Encode text features with dummy variables for asset token status, algorithm, and proof type using pandas get_dummies, then merge with the dataset for modeling.

Learn how principal component analysis enables dimensionality reduction by extracting the most variance-rich components, reducing noise and overfitting to improve generalization and data efficiency.

Scale data with a standard scaler, reduce dimensionality using principal component analysis to three components, and prepare PCA results for clustering with k-means.

Explore the theory of k means clustering, an unsupervised learning method that groups data into k clusters using centroid distance and iterative refinement.

Apply k-means clustering to cryptocurrency data using PCA-reduced features and the elbow method to determine the optimal number of clusters, using inertia to confirm four clusters.

Apply k-means with the optimal four clusters from the elbow curve, train on the x PCA dataframe, and produce a final dataframe with cluster labels for visualization.

Visualize crypto clusters in a three-dimensional scatter plot using principal components, color and shape by cluster, and hover coin names to interpret the results.

Discover how K nearest neighbors, a nonparametric, instance-based model for classification and regression, uses a tunable K to balance flexibility and accuracy. Use cross-validation to select the best K.

Use the cannon or cannabis neighbors classifier to predict whether Dogecoin prices go up or down, by scraping data with the Yahoo! Finance API and preparing a pandas dataframe.

Create price increase/decrease label for crypto data using high, low, and close as features, split for training/testing, scale with StandardScaler, and train a k-nearest neighbors classifier to predict tomorrow's move.

Build and evaluate a k-nearest neighbors classifier using scaled training data, test predictions, and a confusion matrix to compare accuracy and select the best k from 1 to 10.

Learn to find the best k for a k-nearest neighbors model by calculating errors across k values, fitting and predicting, and identifying 14 as optimal for Dogecoin data.

Learn how radius neighbors machine learning uses all samples within a radius to predict cryptocurrency stock movements, extending k-nearest neighbors to regression and density-aware classification.

Load stock data from Yahoo Finance API in Colab using yfinance, covering 2015 to 2021. Create features like returns and buy or sell signals, and save data as a CSV.

Select numerical labels and two feature columns, dropping nulls. Split data into training and testing sets to prepare X and Y for a radius neighbors regression model.

Build a radius neighbors model to predict stock returns, train with radius 1.0, and note a 0.27 score. Use a numpy loop to test radii 0.1–1.5 and compare scores.

Explore CatBoost, an open-source gradient boosting classifier that handles categorical data and integrates with TensorFlow and Core ML, enabling accurate stock buy/sell predictions with minimal training.

Explore gradient boosting as an ensemble learning method using weak learners, typically decision trees, to create strong predictive models for classification and regression, with Cat Boost for crypto decisions.

Load crypto data in Google Colab with yfinance, download XRP USD data (2014–2020), save as CSV, and build a simple buy-or-sell classifier using future prices and a train-test split.

Create a binary buy/sell column from the closing price, prepare X with high, low, close, and volume, and split data into training and testing sets for model training.

Build a CatBoost classifier model on XRP crypto data in Colab, train with iterations and learning rate, and evaluate accuracy to detect overfitting and tune parameters.

Load crypto stock data in Colab with Yahoo Finance API using yfinance, save as CSV, and set open, high, low, volume as features with adjusted close as target for XGBoost.

Build an XGBoost regression model for crypto price prediction, select squared error, assess feature importance, apply simple imputation, and evaluate with mean absolute error and model score.

Discover how deep learning, a multi-layer neural network inside artificial intelligence, uses dense connections, large data, and advanced algorithms to outperform traditional models with ample data.

Explore how artificial neural networks mimic brain learning by layering input, hidden, and output layers, adjusting perceptron weights with activation functions during training across feedforward, convolutional, and recurrent models.