FA-0727Data & Analytics

Data Science with KNIME for Experienced Users

Model selection, evaluation and reusable analytical workflows

Introduction

Why this course

Extend existing KNIME workflow skills with selected machine-learning examples and systematic evaluation. Review preparation, model goals, settings and comparison criteria using supplied datasets and compatible nodes.

The two-day workshop combines a small implementation exercise with broader algorithm demonstrations. It is not a beginner KNIME course or a promise to optimise every listed model in depth.

Learning outcomes

Build on existing workflows to acquire and prepare data for modelling.
Select an appropriate regression, classification or clustering goal.
Compare selected learner settings and features using separate evaluation data.
Interpret model measures and distinguish tuning from independent validation.
Inspect supported model-exchange options and retain reproducible workflow evidence.

Prerequisites

Basic education in mathematics (understanding of linear and non-linear mathematics at high-school level)

Very basic understanding of API and web requests

Experience and understanding of filing system

Experience in Using KNIME for analytics

A compatible KNIME Analytics Platform installation with the extensions and supplied workflows needed for the examples.

Remote participants need reliable connectivity and conferencing tools; a second screen is optional and certification attendance rules are not implied.

Training outline

2 modules

01Day 1 — Preparation, regression and evaluation5 topics

Review node/workflow execution, automation and small flat-file or approved API data inputs.
Explore a supplied iterative/recursive workflow and historical data example where useful; preserve time order when it affects evaluation.
Machine-learning goals, training/evaluation separation and the effect of preparation decisions.
Implement a selected linear-regression or regression-tree workflow; compare polynomial-regression behaviour through a worked example.
Inspect parameters and features using appropriate validation; do not use held-out evaluation data to optimise the workflow.

02Day 2 — Classification, clustering and model exchange6 topics

Compare naïve Bayes, decision trees, k-nearest neighbours, SVM and logistic regression conceptually; practise selected examples rather than five full deep-dive labs.
Review distance metrics, evaluation criteria and algorithm suitability for the sample data.
Clustering concepts: k-means and its cluster-count choice, hierarchical approaches and DBSCAN; compare assumptions and selected results.
Use a bounded parameter-search demonstration; choosing k depends on evidence and purpose, not a guaranteed best grid result.
Inspect PMML/model interchange where supported by the selected nodes; validate input schema and compatibility on reload or scoring.
Assess the workflow, document limitations and identify further practical study.

A programme built around your team.

Share your training goals and requirements.