Applied Machine Learning with BigQuery on Google's Cloud

Learn how BigQuery enables scalable analysis of unstructured data in the cloud, via a web UI for queries, tables, and datasets, with parallel execution.

Explore how Google built BigQuery with Dremel to enable interactive analytics on massive data, achieving fast queries by parallelizing across thousands of machines while using SQL.

Master how BigQuery stores structured data, including fields with names and data types, primitive and nested records, and native, external, and view tables within data sets and projects.

Learn how BigQuery runs queries in parallel across thousands of cores, enabling pipelined execution and high utilization through multi-tenant workloads.

Navigate the Google Cloud Console's BigQuery interface to view and manage datasets, tables, views, and jobs, then create and modify schemas and export data.

Explore BigQuery’s layered architecture with Megastore, Bigtable, and Colossus, featuring columnar storage, column IO, and data replication across data centers for scalable, parallel queries.

Explore how applied machine learning centers on supervised learning, data wrangling processes, and real-world projects, preparing data scientists, machine learning engineers, and data engineers for in-demand roles using Python.

Explore the three core careers in data science—data scientist, machine learning engineer, and data engineer—highlighting their roles, skills, and how they differ from data analysts.

Explore applied machine learning with real-world, supervised models using structured relational data. Compare traditional models like gridiron boosters to neural networks for highly structured data.

Trace the machine learning process from data sourcing and wrangling to modeling, testing on fresh data, and production deployment, with emphasis on supervised learning and Netflix-style recommendations.

Explore the three types of machine learning: supervised, unsupervised, and reinforcement, and their models, including classification and regression, with concepts like binomial and multiclass classification and clustering.

Discover why Python is king by exploring its high-level, human-readable design, open-source nature, and interpreter-driven execution that make Python ideal for data science and machine learning.

Python is open source and free; install the 3.x Anaconda distribution on Windows 64-bit by downloading the installer and selecting a default path to complete the setup.

Install Python on a Mac using the Anaconda distribution, choosing the 3.7 version, opening the Anaconda Navigator, and launching a Python 3 notebook with pandas.

Discover how arrays serve as the core data object in machine learning, from one- and two-dimensional to multidimensional arrays, with CSV data and a Rubik's cube analogy, plus Tensas Flow.

Navigate and manage a Jupyter notebook for machine learning workflows by launching the notebook, creating and renaming notebooks, running code with pandas, saving, editing cells, and using markdown.

Master python libraries for machine learning, including pandas for data massage, exploratory data analysis, and visualizations; source data from SQL Server to create a view and export to CSP file.

Explore core machine learning libraries in Python, including Pandas, NumPy, Matplotlib, and scikit-learn. Import and apply these tools to wrangle data, perform scientific computing, visualize results, and build models.

The lesson introduces core machine learning libraries, including pandas, and demonstrates a Titanic dataset workflow. It covers encoding, binary classification, train-test split, a decision tree classifier, and accuracy evaluation.

Source data from relational databases using SQL Server Management Studio to build a dataset. Join tables to create attributes like age, income, and education, then export as csv for sharing.

Explore exploratory data analysis with pandas, matplotlib, and seaborn on the Titanic dataset; learn to handle missing values, create histograms and bar plots, and compare survival by gender and class.

Apply data cleansing in the machine learning pipeline by selecting essential attributes, handling missing values with imputation, removing noise and outliers, and transforming categorical data to numeric.

Explore supervised learning with gradient boosting by building a binary classifier using XGBoost, pandas, and train-test split, then evaluate accuracy and save the model.

Explore supervised learning fundamentals, including linear regression and binary classification, and distinguish artificial neural networks from deep learning. Understand how regression lines model relationships and how binary outcomes drive decisions.

Master linear regression, a supervised method that predicts an output from a line using two variables, plotting a trend line to forecast sales from temperature in this course.

Demonstrate linear regression to predict scores from study hours in Python, using pandas and matplotlib; split data for training and testing, fit the model, and compare actual versus predicted scores.

Classification is a supervised machine learning technique that separates observations into groups based on characteristics, illustrated by spam detection and Titanic's binary survived versus not survived.

Build a highly accurate classification model using the iris dataset, from importing libraries and encoding labels to splitting data, training a random forest classifier, and achieving 97% accuracy.

Explore the basics of artificial neural networks, including input, hidden, and output layers, and how weights, biases, and activation functions enable learning and decision making.

Explore how to manage datasets and tables, upload and query data, wrangle large data with BigQuery, and scale machine learning in the cloud using Cloud DataLab and GCP tools.

Create datasets and add tables in BigQuery, specifying the dataset location for compliance. Choose data sources, including external federated sources and local uploads, to manage and access your data.

Upload a data set to BigQuery, create a table under the Titanic project with headers, auto-detect the schema from the CSV and skip the first row, then query the table.

Explore Google's Cloud Data Lab, a jupiter notebook–based VM on GCP that you access via cloud shell, spin up with a persistent disk, and preview at port 1881.

Demonstrates modeling the Titanic dataset in Google Cloud DataLab by importing from BigQuery, building a dataframe, splitting data, and training a decision tree classifier with metrics.

Model the iris dataset on Google Cloud Datalab by uploading a notebook, importing the data, and training a random forest classifier with a train/test split in a Jupyter notebook.

Upgrade your cloud DataLab by resizing the instance to more resources, then start and connect to the DataLab for scalable model building.

Discover how BigQuery ML democratizes machine learning by letting analysts build and evaluate linear regression, binary logistic regression, and multiclass logistic regression models directly in BigQuery, without moving data.

Create a binary logistic regression model in BigQuery using the Titanic view and define the input label. Evaluate the model with accuracy 77 percent and deploy predictions on fresh data.

Install and configure the Google Cloud SDK on Windows, resolve a firewall sign-in issue, authenticate, and set up the default project to start using gcloud.

Explore gsutil navigation basics in Google Cloud Storage: create and manage buckets, upload and download files, set access control lists, grant public access, and perform parallel transfers.

Segment and move large datasets to GCP by exporting two SQL Server query results to separate files, uploading to Cloud Storage, and joining them in BigQuery with a union.