Building Machine Learning Models for Predictions and Patterns

• Machine learning and its relationship to AI
• Real-world ML applications
• Training data and pattern recognition
• Features and target variables
• Data collection
• Preparing data before training a model

• Supervised machine learning
• Labeled datasets
• Classification
• Classification algorithms
• Model performance metrics
• Confusion matrices
• Regression
• Predicting numerical outcomes
• Forecasting values over time

Hands-on Labs: Train Your First Prediction Model — Train competing classification models and select one based on measured results. Build a Regression Model — Train a regression model, compare predicted and actual values, calculate error, and create a forecast.

• Unsupervised learning
• Clustering
• Similarity and distance
• Interpreting clusters
• Customer and operational segmentation
• Anomaly detection
• Fraud and security use cases

Hands-on Lab: Find Customer Segments and Anomalies — Cluster a business dataset into meaningful groups, interpret each segment, detect unusual records, and review whether identified anomalies represent errors or significant events.

• Training versus testing data
• Validation sets
• Overfitting
• Underfitting
• Comparing training and test performance
• False positives and false negatives
• Threshold selection
• Interpreting model errors

Hands-on Lab: Split, Test, and Diagnose — Split a dataset, evaluate model performance on unseen records, identify signs of overfitting or underfitting, and select an operating threshold based on error impact.

• Privacy-preserving machine learning
• Differential privacy
• Federated learning
• Balancing privacy and model performance
• Selecting privacy controls based on data sensitivity

• Neural-network structure
• Layers and learned representations
• Deep learning
• When neural networks are appropriate
• Traditional ML versus deep learning
• Image-classification applications

Hands-on Lab: Train an Image Classifier — Train and test an image classifier using a browser-based machine-learning tool, measure performance against held-out images, and improve the model through additional training examples.

• Customer segmentation
• Lead scoring
• Revenue and demand forecasting
• Fraud detection
• Risk assessment
• Supply-chain applications
• Operational prediction
• Moving a model from experiment toward production use

Choose a defined business problem and build a working predictive solution using an appropriate classification, regression, forecasting, clustering, or anomaly-detection approach. Submit the model workflow, performance evidence, interpretation of the results, and a concise model summary explaining how the prediction should be used in the business process. The capstone is assessed against a skills-based rubric and must be successfully completed to earn the CERT4TECH Certificate in AI Predictions and Pattern Finding.