FE590 Introduction to Knowledge Engineering

From Hanlon Financial Systems Lab Web Encyclopedia
Jump to: navigation, search



Course Catalog Description

Introduction

Introduction to information theory: the thermodynamic approach of Shannon and Brillouin. Data conditioning, model dissection, extrapolation, and other issues in building industrial strength data-driven models. Pattern recognition-based modeling and data mining: theory and algorithmic structure of clustering, classi cation, feature extraction, Radial Basis Functions, and other data mining techniques. Non-linear data-driven model building through pattern identi cation and knowledge extraction. Adaptive learning systems and genetic algorithms. Case studies emphasizing nancial applications: handling nancial, economic, market, and demographic data; and time series analysis and leading indicator identi cation.
Campus Fall Spring Summer
On Campus X X
Web Campus X X

Instructors

Professor Email Office
Thomas Lonon
tlonon@stevens.edu Altofer 301



More Information

Course Description

Overview

  • This course provides an applied overview of both classical linear approaches to statistical learning and more modern statistical methods.
  • The classical linear approaches will include logistic regression, linear discriminant analysis, k-means clustering, and nearest neighbors.
  • The more modern approaches will include generalized additive models, decision trees, boosting, bagging, support vector machines, and others.

Prerequisites:Knowledge of R (or willingness to learn) Prob/Stat background

Course Outcomes

At the end of this course, students will be able to:

1. Describe the difference between supervised/unsupervised learning and parametric/nonparametric models

2. List a variety of techniques for each type of model from above


3. Apply the various techniques to sets of data

4. Interpret which model seems to fit the data set the most productively


Course Resources

Textbook

Additional References

Whatever you feel is best. If you need help with any of the background, feel free to reach out with questions.



Grading

Grading Policies

Grades will be based on a combination of quizzes and assignments.

1. Quizzes. There will be two multiple choice online quizzes.

2. Assignments. There will be four homework assignments in which students will write programs to solve problems related to statistical learning.

Weights.

In computing the course grade, each activity will be assigned a weight (subject to the instructor's discretion) as follows:

Item Weight

Quizzes 20%

Assignment 1 20%

Assignment 2 20%

Assignment 3 20%

Assignment 4 20%

Extra Credit

There are no "extra assignments" that students can do to raise their average outside of the ones assigned. There are no exceptions, don't even bother coming to me and asking about extra work and the end of the semester, as I will only direct your attention to this part of the syllabus.


Lecture Outline

Topic Reading
Week 1 Overview of Supervised Learning
Week 2 Linear Methods for Regression
Week 3 Linear Methods for Regression
Week 4 Linear Methods for Classification
Week 5 Basic Expansions and Regularization
Week 6 Basic Expansions and Regularization
Week 7 Model Assessment and Selection
Week 8 Model Inference and Averaging
Week 9 Model Inference and Averaging
Week 10 Additive Models, Tree Related Methods
Week 11 Additive Models, Tree Related Methods
Week 12 Boosting and Additive Trees
Week 13 Support Vector Machines
Week 14 Unsupervised Learning