ThingWorx Analytics Training: Module 3 Part 3
- December 23, 2022
- 0 replies
- 921 views
This video concludes Module 3: Data Profiling of the ThingWorx Analytics Training videos. It shows you a few examples of questions that should be asked of a subject-matter expert (SME) to better understand the information contained in a dataset. Using answers to these questions, you will use a tool such as Microsoft Excel to modify a given dataset, and prepare it for future exercises in this course.
This video contains hands-on exercises, and it is recommended to complete them before moving on to subsequent sections. Download and use the attached "BEAN_PRO_DATA_UNPROCESSED.csv" for the exercises. To compare your work against a processed version of the data, see "BEAN_PRO_DATA_PROCESSED.csv"
This video has chapters — use the chapter menu in the player to jump to any section.
Chapter Summaries
0:00 – Introduction: BeanPro Espresso Machine Dataset & Data Profiling Results Introduces the data profiling results for BeanPro, a fictitious espresso machine manufacturer investigating machine failures, and explains how the profiling document surfaces data quality issues before analysis.
0:26 – Profiling Findings: Duplicate Records & Toronto-to-Canada Region Consistency Reviews the first profiling findings — duplicate records that must be removed and a Region column where the city "Toronto" should be replaced with "Canada" to keep regional values consistent.
0:48 – Recommended Data Fixes: Missing Priority Codes, Alert Code Grouping, Train/Score Split & Sparse Columns Summarizes the remaining recommendations: imputing missing Priority Codes Levels 0–2 (Prior 1–30 days) with zeros per SME guidance, regrouping User Alert Codes 6–14 into alert code 2, adding a training/scoring split feature, and dropping the sparse Replaced Parts 30/60/90-day columns.
1:41 – Applying Corrective Actions: Region Cleanup (Toronto → Canada) Begins the hands-on cleanup by replacing every record with the Toronto region value and reassigning it to the correct Canada region before loading the data into ThingWorx Analytics.
2:12 – Imputing Missing Priority Code Values with Zeros Demonstrates filling missing values in the Priority Codes Levels 0–2 Prior 1–30 days column with zeros, since gaps here would degrade the quality of the predictive model.
2:45 – Formula-Based Alert Code Grouping (Values >5 → Code 2) Shows how to create a new formula-driven column that replaces any User Alert Code above five with the value 2, then copies those results back over the original alert code values.
3:53 – Feature Engineering: Creating the Training/Scoring Split Column Walks through feature engineering to add a new column that will later be used to divide the dataset into separate training and scoring subsets for model building.
4:33 – Removing Sparse Columns: Replaced Parts 30/60/90-Day Fields Explains why columns with excessive missing data hurt machine learning model quality and removes the Replaced Parts in Last 30, 60, and 90 Days columns entirely; small-gap handling is deferred to a later module.
5:10 – Removing Duplicate Records & Finalizing the Analytics-Ready Dataset Completes preparation by deleting duplicate records, producing a clean, analytics-ready BeanPro dataset that is ready to load into ThingWorx Analytics.
5:54 – Module Wrap-Up: Upcoming Modules & PTC Customer Success Contact Closes the ThingWorx Analytics training module, previews upcoming modules in the program, and points viewers to a PTC customer success representative for more information.