Outlier Detection Algorithms in Data Mining and Data Science

Course details

Welcome to the course " Outlier Detection Techniques ". 

The process of identifying outliers has many names in Data Mining and Machine learning such as outlier mining, outlier modeling and novelty detection and anomaly detection.

Despite the importance and huge academic publications on the issue, " Outlier Detection Techniques " is not popular enough.

In my opinion, Everyone, who deals with the data and analyze it,  needs to know  "Outlier Detection Techniques". So, this course  is unique.

Outlier detection algorithms are useful in areas such as: Data Mining, Machine Learning, Data Science, Pattern Recognition, Data Cleansing, Data Warehousing, Data Analysis, and Statistics. In this course, you will learn approaches that come from Statistics discipline and Data Mining. 

I will present you on the one hand, very popular algorithms used in industry, but on the other hand, i will introduce you also new and advanced methods developed in recent years, coming from Data Mining.

You will learn four algorithms for detection outliers in Univariate space, two algorithms in multivariate space and also learn innovative algorithm for detection outliers in high-dimensional space.

I am convinced that only those who are familiar with the details of the methodology and know all the stages of the calculation, can understand it in depth. So, in my teaching method, I put a stronger emphasis on understanding the material, and less on programming. However, anyone who interested in programming, I developed all algorithms in R,  so you can download and run them.

In Python,  you can use scikit-learn library, that  provides a set of Machine Learning tools that can be used in outlier detection.

In R, you can use different libraries, such as robustbase (Adjusted Boxplot Rule), DMwR (LOF algorithm), HighDimOut (ABOD algorithm) and et cetera. 

You will be impress, that  Outlier Detection it is a world unto itself. It is a complex but very interesting world.

The course is suitable for everyone, even if you have no background in statistics and linear algebra. The most important thing is desire to learn.

At the end of the course, you will not only be familiar with a variety of "Outlier Detection Techniques", but also know to implement them. During the course, i put a strong emphasis on practice(including quizzes) to make sure you understand each topic.

List of Algorithms:

Univariate space:

1. Three Sigma Rule ( Statistics +  R  code)

2. MAD ( Statistics +  R  code )

3. Boxplot Rule ( Statistics +  R  code )

4. Adjusted Boxplot Rule ( Statistics +  R  code )

Multivariate Space :

5. Mahalanobis Rule ( Statistics +  R  code )

6. LOF - Local Outlier Factor ( Data Mining +  R  code)

High-dimensional Space:

7. ABOD - Angle-Based Outlier Detection ( Data Mining +  R  code)

I sincerely hope you will enjoy the course.