D603 Machine Learning is a three-unit course in the third term of the Data Science specialization of WGU's Master of Science in Data Analytics.
WGU describes machine learning as the broad discipline of developing algorithms and statistical models to predict, classify or cluster data and that improve over time.
The course focuses on building, training, running and testing supervised and unsupervised models and quantifying their accuracy and precision to decide which suits a business situation. The listed prerequisites are Analytics Programming and Statistical Data Mining. Students often find that choosing and justifying a model is harder than fitting one.
Course at a Glance
| Item | Details |
|---|---|
| University | Western Governors University |
| Course code | D603 |
| Level | Graduate, Data Science specialization |
| Prerequisites | Analytics Programming, Statistical Data Mining |
| Standard path | Term 3, 3 competency units |
| Typical work | Model comparison for business problems and forecasting |
What D603 Covers
WGU lists supervised methods including k-nearest neighbors, logistic regression, decision trees and support vector machines, unsupervised methods including k-means clustering, hierarchical clustering and t-SNE, and ensemble methods.
| Competency (WGU wording, summarized) | What it means in practice |
|---|---|
| Recommends a supervised model based on comparison of performance for a business problem | Fitting several models and choosing with evidence |
| Recommends an unsupervised model based on comparison of performance for a business problem | Comparing clustering or dimension reduction approaches |
| Applies time series models to generate forecasts | Modeling trend and seasonality to predict ahead |
Key Concepts Explained
Training and Testing
Split data so models are judged on data they have not seen. Evaluating only on training data overstates performance.
Illustration (hypothetical): A decision tree scores very high on its training data but much lower on held-out data. That gap signals overfitting, and a simpler tree or an ensemble may generalize better. Reporting both scores shows the reasoning.
Choosing Performance Measures
Accuracy can mislead when classes are imbalanced. Consider precision, recall and related measures, and choose according to the cost of each type of error in the business case.
Time Series
Time series models use the ordering of observations. Check for trend and seasonality, split by time rather than at random, and judge forecasts against held-out later periods.
Typical Assignments and How to Approach Them
| Task type | What it tests | How to approach it |
|---|---|---|
| Supervised model comparison | Evaluation and judgment | Use the same data split and metrics for each model |
| Unsupervised model comparison | Interpretation | Explain how you judged cluster quality |
| Forecasting | Time series skill | Test on later data and report error |
A Fair Model Comparison
Comparison only means something if conditions are equal. Use the same train and test split, the same preprocessing and the same evaluation measures for every model, then tabulate results.
| Model | Strength to mention | Limitation to mention |
|---|---|---|
| k-nearest neighbors | Simple, few assumptions | Slow on large data, sensitive to scale |
| Logistic regression | Interpretable | Assumes a linear relationship in log-odds |
| Decision tree | Easy to explain | Can overfit |
| Support vector machine | Effective in many settings | Harder to interpret and tune |
Judging Unsupervised Models
Without labels there is no single accuracy score. Use measures of cluster quality, look at whether clusters make business sense, and test stability. Explain what each cluster represents so the result can inform a decision, such as customer segments for targeted marketing.
Self-Check Before You Submit
- Did you use the same split and metrics for every model?
- Have you explained why your chosen measure fits the business cost of errors?
- For forecasting, did you split by time and report forecast error?
- Does your recommendation explain the trade-off between accuracy and interpretability?
Connecting the Maths to the Business
Evaluators look for a clear line from metric to decision. If missing a fraudulent transaction costs far more than flagging a genuine one, recall may matter more than overall accuracy.
State the cost trade-off in plain words, choose the measure that reflects it, and show how the final model performs on that measure. That link is what turns a technical comparison into a recommendation.
Where Students Get Stuck
- Data leakage. Keep test data separate through preprocessing.
- Metric choice. Link measures to business costs.
- Black-box answers. Explain why a model fits.
- Time series splits. Do not shuffle time-ordered data.
Study Tips for D603
- Standardize your workflow so every model gets the same treatment.
- Record settings and results in a table.
- Write the business recommendation in plain language.
- Review D600 regression and PCA notes.
How We Help with D603
Send the prompt, rubric and data. We can prepare a custom analysis, explain algorithms, or review your comparison. See our regression analysis help guide.
GradeEssays is independent and not affiliated with WGU. Our work is a study and reference aid. Your submission must be your own work under WGU's academic integrity policy. Orders are written from scratch, with free revisions within the scope of your original request and a full refund if late.
Get Your Model Comparison Right
Share the prompt, rubric and data. We prepare a custom analysis or review your work.
Start My D603 HelpFree revisions · Full refund if late · Written from scratch for your order
Frequently Asked Questions
Supervised and unsupervised machine learning, ensemble methods and time series forecasting.
Data Science, within the MS Data Analytics.
Analytics Programming and Statistical Data Mining.
k-nearest neighbors, logistic regression, decision trees, support vector machines, k-means, hierarchical clustering and t-SNE.
Term three of the Data Science standard path.
No. We provide tutoring, custom answers, editing and feedback only.