ENCE 688P: Data Mining and Machine Learning for the Built Environment
Mark Austin, Department of Civil and Environmental Engineering, University of Maryland, College Park.	CLASS Notes from Class: [ Spring 2022 ] PROJECTS Project Abstracts: [ Fall 2020 ] [ Fall 2021 ] [ Spring 2022 ]
GOALS This course is a hands-on introduction to tools and techniques for data mining and machine learning analysis of data from the built environment. Students will learn how to write and compile simple software programs in Python and Java (no prior experience required), and then work step-by-step toward the design of software solutions using objects, data structures and software patterns. Students will be introduced to a spectrum of software tools, basic methods, and algorithms for data mining (e.g., Weka) and machine learning analysis of data (e.g., DL4J, TensorFlow, Keras). The semester will conclude with completion and presentation of a project. SPRING SEMESTER, 2022 The topics will be as follows: Part 1: Data and Information Management in the Built Environment (2 weeks) Learning about the Built Environment Topic: Modern Infrastructure Systems and Near-Term Challenges. Topic: Engineering Sensor Systems. Topic: Large-scale Urban and Global Sensing. Opportunities for AI, Data Mining and Machine Learning Topic: AI in the 1980s, 1990s, 2000s, and post 2010 era. Topic: Recent Advances in Data Mining/AI/Machine Learning. Topic: Cyber-Physical and Digital Twin Systems. Part 2: Object-Oriented Software Development (3 weeks) Introduction to Engineering Software Development Topic: Evolution of computer languages over the past 50 years? Topic: Low- and high-level languages Topic: Scripting languages versus compiled languages Getting Started with Python Topic: Programming (data types, expressions, assignments, branching constructs, loops), Topic: Arrays and dictionaries Topic: Data visualization (charts) Topic: Python Productivity Tools: Google Colab, Apache Zeppelin Getting Started with Java Topic: Writing and Compiling a Simple Python/Java Program, Topic: Programming (data types, expressions, assignments, branching constructs, loops, methods), Topic: Single- and multi-dimensional arrays, Topic: Java Productivity Tools: Ant Object-based Modeling with Python and Java Topic: Classes, objects, association and inheritance relationships. Topic: Abstract classes and interfaces. Topic: Guidelines for class and package design. Part 3: Collections, Basic Data Structures + Algorithms (2 weeks) Java Collections Framework Topic: Collection, Set and Map Interfaces Topic: Working with Arraylists, Sets and HashMaps. Topic: Basic Algorithms for Searching and Sorting Python Collections Module Topic: Basic Collections: lists, sets, tuples, dictionaries. Topic: Working with lists, sets, tuples and dictionaries. Part 4: Hands-On Data Mining and Machine Learning (5 weeks) Working with Real-World Datasets Topic: Urban, Government and Geographic Data Portals Topic: Data Analysis with Pandas Data Mining Tools and Techniques Topic: Basic Methods (Classification, Association, Clustering). Topic: Synthesis of Decision Trees and Rules Topic: Data Mining with Weka (Applications) Introduction to Machine Learning Techniques and Tools Topic: Machine Learning Capabilities Topic: Taxonomy of Machine Learning Problems Topic: Types of Machine Learning Systems Topic: Urban Applications of Machine Learning Introduction to Neural Networks Topic: Perceptron Models (Building Block of Machine Learning) Topic: Activation Functions, Loss Functions, Metrics of Evaluation Topic: Training Neural Networks (Backpropogation and Optimization Algorithms) Multilayer Neural Networks Topic: Multilayer Network Architectures and Capabilities Topic: Neural Networks with One Hidden Layer Topic: Neural Networks with Two Hidden Layers Hands-On Machine Learning Topic: Software Setup (TensorFlow, Keras, Jupyter, Anaconda, Deep Learning 4 Java) Topic: Tensors and TensorFlow Graphs Topic: Working with TensorFlow and Keras Topic: Working with DL4J (Deep Learning 4 Java) Part 5: Additional Topics (Class Interest and Time Permitting) Software Design Patterns Topic: Design Patterns in Architecture and City Planning. Topic: Software Design Patterns (e.g, builder, composite hierarchy, visitor, model-view-controller). Advanced Machine Machine Learning Topic: AutoEncoder Neural Networks (new for 2021/2022) Topic: Recurrent Neural Networks (new for 2021/2022) Students will complete individual homework assignments, and work in small teams on a data mining/machine learning project.
ENCE 688P ONLINE Here's what online means: There with no in-person contact. For each lecture I will post the "lecture content" (pdf) and a "recorded video" (zoom video) to the notes from class page. I will also post handouts and links to interesting web sites on notes from class . Online office hours (zoom) where we can review the material and answer your questions. E-mail submission of homework and project work. Please see the detailed instructions (for naming of files) below. Office Hours / Synchronous Class Sessions Mark Austin . Office hours and synchronous class sessions will be Monday and Wednesday at 5-6 pm. Join Zoom Meeting: https://umd.zoom.us/j/6517468335 Even if you just want to drop-in to catch up, that'll be fine too! If 5-6 pm doesn't work for you, send an e-mail (austin "at" umd.edu) and we will work something out. Submission of Homework and Project Work Homework will be posted on the notes from class web page. Please submit your homework as a zip file and send either as an attachment to an e-mail or via Dropbox. Also, please indicate in your e-mail subject heading the class and purpose of the e-mail, e.g,, ENCE688P: Homework 2 ... Class Text and Resources Text not required, but there will be lots of class handouts on data mining and machine learning. There is a great data mining text: Witten I.H., Frank E., Hall M.A., and Pal C.J., Data Mining: Practical Machine Learning Tools and Techniques, Fourth Edition, Morgan Kaufman, 2017. which you might consider getting -- well worth the money! Java and Python software will be distributed via Dropbox. Course Assessment and Exam Schedule The course will be assessment will be as follows: Homework (30%). Mid-semester project proposal and presentation (20%). End-of-semester project/report in data mining/machine learning (50%). Note. No mid-term exams. No final exam. Let's focus on making fabulous projects! Accommodation for students with disabilities will be made. At the end of the semester, please participate in the evaluation of courses through CourseEvalUM. Your feedback is confidential and an important means of improving the course in future semesters.
Download Python and Java Download Python 3.X . It seems that Apple ships its Macs with Python 2.X pre-installed. But for the purposes of this class I am going to assume you have Python 3.7 (or Python 3.8) installed. This detail matters because the language is not backwards compatible (Strike 1 against Python). Download Java . I have Java 1.11 on my laptop, but you are certainly welcome to download a more recent version. Download Apache Ant . Apache Ant is a Java Library that manages the compilation of Java programs and execution of programs and test cases. Extremely useful. Click here to download the Eclipse IDE. If your computer is a Mac think about downloading Homebrew and then using brew to automatically download and install the Ant packages on your machine. Note (Oct, 2020). I just installed homebrew on my iMac (home) running Catalina (10.15) and, sadly, Apple has not made this process easy. Here's the bottom line to get things working: Homebrew uses something called the "command line tools", which is an addition to Xcode. So before you can install Homebrew, you need to download and install the command line tools. I downloaded the dmg file for the command line tools from the Apple developers web site -- just create an account, it's free. Run the installation program and the tools will be put in /Library/Developers/CommandLineTools/ Then, to install Homebrew, cut-and-paste the command listed on the Homebrew page into a terminal window running the bash shell. The installation only takes are few minutes and you are good to go. For example, to install ant, simply type: brew install ant at the prompt in a terminal window. Python and Java Programming Resources Python 3.8 Tutorial The Python Collections Module Shapely : Python package for manipulation and analysis of planar geometric objects. Also see Shapely User Manual . Java Programming Source Code Examples (www.java2s.com) Standard and Extension Packages (i.e., packages java.* and javax.*) in Java 8. Java Language Specification SE8 Release 2015 ( pdf ) This document contains a detailed description of Java SE8 (handy is you want to know exactly what Java 8 does). Standard and Extension Packages (i.e., packages java.* and javax.*) in Java 6 ( pdf ). List of Java Frameworks Download JAMA: A Matrix Package. JAMA is a basic linear algebra package for Java. AI and ML Software Artificial Intelligence: A Modern Approach (AIMA). Code website on Github. Apache OpenNLP : A Machine Learning-based Toolkit for the processing of Natural Language Text. Written completely in Java. Source code: github . Weka Workbench for Machine Learning Miscellaneous Deep Learning for Java (DL4J) Library Top 69 DL4J Open Source Projects on Github. Smile: Statistical Machine Intelligence and Learning Engine Google Cloud AutoML (Automated Machine Learning). See the Wikipedia page for a general description of AutoML goals and capabilities. Real-World Datasets The 50 Best Free Datasets for Machine Learning Best Public Datasets for Machine Learning and Data Science Urban datasets on data.world PEER Hub ImageNet Open Data DC, District of Columbia. Open Data Portal , District of Columbia Water and Sewer Authority Washington Metropolitan Area Transit Authority API. New York City OpenData Portal. City of Chicago Data Portal . Chicago: Array of Things project and data. Kaggle: Large collection of public datasets and notebooks. UCI Machine Learning Repository . Working with Real-World Data Data Analysis with Pandas . Real-world data sources are nearly always very messy. Pandas is an open source data analysis and manipulation tool, built on top of Python. Click here to install pandas 1.1.4 (Released October, 2020). Python Pandas Tutorial at TutorialsPoint.com Data Science, Data Mining, Neural Networks Weka: Attribute-Relation File Format (ARFF) Click here for online coversion of CSV files to ARFF format. Weka Manual Version 3-8-3 (pdf), University of Waikato, 2018. Weka Workbench Appendix (pdf), 2016. See the Tutorial on Weka YouTube videoes for Data Mining with Weka MOOC . Gilbert Strang (MIT Opencourseware) on Matrix Methods in Data Analysis In my view, Gilbert is a rock star. YouTube Video by Dan Roberts, MIT and Salesforce, The Principles of Deep Learning Theory, IAS Physics Group Meeting, October 2021. Entropy: How Decision Trees make Decisions Tutorial: Main Types of Neural Networks and Applications Comprehensive Guide to Types of Neural Networks Building Neural Networks in Python TensorFlow Core . Introduction to TensorFlow 2, Basics and Keras. TensorFlow Tutorial at TutorialsPoint. TensorFlow: Programming with Data Flow Graphs Apps and TensorFlow: Swift for TensorFlow Keras: The Python Deep Learning API Keras Cheat Sheet (pdf). PerceptiLabs: TensorFlow-based Visual Machine Learning Modeling, 2020. CS 231n: Convolution Neural Networks for Visual Recognition , Stanford University, Spring 2020. CS 224W: Machine Learning with Graphs , Stanford University, Winter 2021. CS 11-785: Introduction to Deep Learning , Carnegie Mellon University, Spring 2021. Jason Brownlee, Tour of Machine Learning Algorithms , 2018-2020. Artificial Intelligence: A Modern Approach. Fantastic resource for AI knowledge and software. Part V of the new edition is devoted to machine learning. Digital Twins National Digital Twin Program, Centre for Digital Built Britain, University of Cambridge, England. See article: Data Science and Computational Mathematics to Advance Predictive Engineering . Digital Twin Cloud Services for Infrastructure Engineering , Bentley Systems, 2021. Making Digital Twins to Monitor and Control Tomorrow's Reactor Designs, Idaho National Library, Idaho Falls. Also see DICE, their Digital Innovation Center of Excellence. Here's an article on construction and operation of their hydrogen hub prototype (December, 2022). Corral-Acero J., et al., Digital Twins and Vision of Precision Cardiology , European Heart Journal, Volume 41, Issue 48, 21 December 2020, pp. 4556--4564. Inter-Professional Projects Program: Creating an IIT Digital Twin , IIT, Chicago, Fall Semester 2021. Students uploaded videos of their point cloud models to YouTube. Here is a really good example . Algorithms and Software for Anomaly Detection Anomaly Detection, Wikipedia. Modern Recipes for Anomaly Detection Time Series Anomaly Detection using an AutoEncoder, Keras, 2020. Deep Learning for Anomaly Detection, Cloudera, 2020. AutoEncoders and the Denoising Feature: From Theory to Practice , Towards Data Science, 2020. Time Series Darts: Time Series Made Easy in Python. Big Data Algorithms + Tutorials (useful techniques that are beyond this course) Sketching Algorithms for Big Data Data Streaming and Compressed Sensing. Miscellaneous Real-World Applications and Resources Bloomberg City Lab, How Cities are using Digital Twins like SimCity for Policymakers , Bloomberg, April 5, 2022. Centre for Smart Infrastructure and Construction, University of Cambridge, UK. Thorton Tomasetti Core Studio Design (Intersection of Design, Computation, Research) OpenStreetMap Resources on Github OpenRailwayMap Resources Audio Data Analysis with Python: Part 01. Using ML to Improve Flow of Traffic at Intersections , Texas A&M, 2021. Deepmind uses ML to improve the efficiency of matrix multiplication.
Developed in August 2020 by Mark Austin Copyright © 2020-2022, Department of Civil and Environmental Engineering, University of Maryland