Nirpy Webinar
About
Title: Class modelling and outlier detection in NIR spectroscopyPresenter: Daniel Pelliccia
Abstract
Classification is a fundamental task in multivariate data analysis. Conventionally, one can distinguish between two distinct approaches to the classification problem: discriminant methods and class modelling.
Discriminant methods seek to find optimal boundaries between predefined groups. Classic examples are Linear Discriminant Analysis (LDA) or Partial Least Squared Discriminant Analysis (PLS-DA). These methods are suitable when all the relevant classes are known and represented in the training data.
Conversely, class modelling approaches build descriptive models of individual classes. The classic example of this approach is Soft Independent Modelling of Class Analogy (SIMCA). Given a sample and a class, SIMCA gives a rule to decide whether the sample does or doesn't belong to that class. As such, the result of classification one class, multiple classes, or none at all.
SIMCA models are suitable to tackle 'novelty detection', which is closely related to outlier detection. This tutorial will explore this connection.
The talk will present examples related to the classification of near-infrared spectra using publicly available datasets and examples in Python.
Session Times
Dates and times are displayed in UTC+10 (Melbourne, Australia). To help you with scheduling, these are the session times in a few different times zones.Session 1 (best for the Americas): Jun 16th, 06:00 - 07:00 (UTC+10)
Session 2 (best for EMEA - APAC): Jun 16th, 18:00 - 19:00 (UTC+10)
Dates
Tuesday 16 June 2026 (UTC+10)Location
Online event access details will be provided by the event organiser
Sessions
SESSIONS
DESCRIPTION
STATUS
Contact Form
Nirpy Webinar
Resend recently purchased tickets to our events to your email.
Enter email to start search
You can review the event organiser's terms and conditions here.