iwantcoding.com
🔥 Daily 👥 Rooms 🏆 Top Log in Sign up

9.1 Predictive Analytics & ML

Predictive analytics and machine learning forecast outcomes from historical patterns. The book covers the supervised / unsupervised split, ML pipelines, MLOps, and Philippine industry use cases.

9.1 Predictive Analytics and Machine Learning

Three learning paradigms

ParadigmData neededExample task
SupervisedLabelledClassification + regression
UnsupervisedNo labelsClustering + reduction
ReinforcementReward signalAgent decisions

The ML pipeline

Collect Clean Features Train Evaluate Serve Monitor drift detected -> retrain Figure 9.1 - The ML production loop: collect, train, serve, monitor, retrain.

Common algorithms by task

TaskAlgorithms
ClassificationLogistic, random forest, XGBoost, neural net
RegressionLinear, ridge, gradient boosting, neural net
Clusteringk-means, DBSCAN, hierarchical
RecommendationMatrix factorisation, two-tower neural
ForecastingARIMA, Prophet, recurrent / temporal nets

MLOps essentials

ComponentTool examples
VersioningDVC, MLflow
PipelinesVertex AI Pipelines, Kubeflow
Feature storeFeast, SageMaker Feature Store
Model registryMLflow, SageMaker
Drift monitoringEvidently AI, Arize, WhyLabs

Risks the EA must address

RiskMitigation
BiasFairness metrics per protected attribute
DriftScheduled retrain + monitoring
ExplainabilitySHAP / LIME stored per decision
Data leakageTrain/test split discipline
Lineage gapsDVC + metadata catalogue
Mentor’s tip: Pipeline first, model second. MLOps gives you reproducibility, drift detection, and audit. Fair-lending and explainability are statutory in banking; bake them into the pipeline before deploying.

Discussion

Loading…