Machine Learning with Python

Programme Summary

Machine learning projects require careful problem definition, trustworthy data and honest evaluation. This programme develops practical skills for preparing tabular data, training supervised models and comparing performance in Python. Participants build a reproducible modelling workflow, control data leakage and document limitations, then explain how a selected model should be reviewed before any operational use.

Training Objectives

1 Develop sound problem framing for machine learning; 2 Strengthen data preparation and validation discipline; 3 Build practical supervised modelling and evaluation skills; 4 Support reproducible workflows and responsible interpretation; 5 Connect model results with business decision requirements.

Learning Outcomes

1 Frame a prediction problem and evaluation criteria; 2 Prepare data without leaking test information; 3 Train and compare suitable baseline models; 4 Interpret performance uncertainty and subgroup limitations; 5 Deliver a reproducible model evaluation report.

Target Audience

Data Analysts; Business Intelligence Analysts; Junior Data Scientists; Analytics Engineers; Technical Business Analysts.

Target Industries

Retail Analytics; Manufacturing Analytics; Financial Services Analytics; Logistics Analytics; Telecommunications; Business Technology.

Course Contents

• Module 1 Machine Learning Problem Framing: Prediction tasks and business decisions; Classification regression and unsupervised distinctions; Success measures and nonlearning alternatives.
• Module 2 Python Modelling Environment: Notebooks libraries and reproducible environments; Data structures functions and workflow organisation; Random seeds versions and experiment records.
• Module 3 Data Understanding: Target definition features and observation units; Data provenance permissions and representativeness; Missing values outliers and sampling limitations.
• Module 4 Exploratory Analysis: Distributions relationships and class imbalance; Visual checks and descriptive statistics; Identifying anomalies without premature conclusions.
• Module 5 Training Validation and Test Sets: Independent evaluation and holdout design; Time group and stratified splitting; Preventing leakage across data partitions.
• Module 6 Data Preparation Pipelines: Imputation encoding and scaling decisions; Fitting transformations within training data; Reusable pipelines and consistent inference preparation.
• Module 7 Baseline Models: Simple rules and naive comparison models; Establishing a meaningful performance benchmark; Understanding when complexity adds little value.
• Module 8 Regression Models: Linear regression assumptions and interpretation; Error measures and residual analysis; Regularisation and overfitting control.
• Module 9 Classification Models: Logistic regression and decision boundaries; Probability outputs and threshold decisions; Class weighting and imbalance considerations.
• Module 10 Tree Based Methods: Decision trees and model complexity; Ensemble methods and variance reduction; Comparing performance with interpretable baselines.
• Module 11 Model Evaluation: Confusion matrices precision and recall; Regression errors and business consequences; Selecting metrics appropriate to the decision.
• Module 12 Cross Validation: Fold design and validation consistency; Grouped and time dependent data considerations; Interpreting variability across validation runs.
• Module 13 Hyperparameter Selection: Search strategies and computational budgets; Validation based tuning and overfitting risks; Preserving the final test set.
• Module 14 Feature Engineering: Domain informed transformations and interactions; Feature selection within validation procedures; Avoiding target derived or unavailable features.
• Module 15 Interpretation and Limitations: Coefficients importance and local explanation concepts; Correlated features and causal inference limits; Communicating uncertainty and model weaknesses.
• Module 16 Fairness and Responsible Modelling: Relevant subgroup performance comparisons; Privacy sensitive attributes and justified use; Human review and unacceptable use boundaries.
• Module 17 Unsupervised Learning Overview: Clustering dimensionality reduction and exploration; Scaling distance choices and interpretation; Limits of inferred groups and patterns.
• Module 18 Operational Readiness: Input validation monitoring and drift indicators; Model versioning ownership and rollback concepts; Security review and deployment approval interfaces.
• Module 19 Model Documentation: Problem data methods and evaluation records; Assumptions limitations and intended use; Reproducible outputs and reviewable notebooks.
• Module 20 Modelling Capstone: Building a complete tabular modelling workflow; Comparing models on defensible evaluation evidence; Presenting results limitations and next actions.

Assessment

• 100 single best answer questions, including applied scenarios; one mark per question.
• 150 minute invigilated digital examination with identity verification. Approved accessibility adjustments preserve the assessed competence.
• Supplied reference pack permitted; external internet and AI assistance prohibited.
• Individual applied assessment (90 minutes): Build a supervised learning pipeline using supplied data, compare a baseline and selected model, and submit a reproducible notebook with evaluation, leakage checks and limitations.
• Assessment criteria: Correctness of data splitting and pipeline design; Suitability and integrity of model evaluation; Reproducibility and clarity of limitations.
• Pass each component separately: 75/100 written marks (75%) and at least 70% on the applied task.
• All mandatory criteria must pass: Do not train on heldout test outcomes or conceal evaluation leakage.
• Individual results and feedback are provided. One reassessment may follow coaching under the published enrolment terms.

Certification

• Successful learners receive a TQCSI Arabia digital Certificate of Successful Completion.
• The certificate records the learner name, exact programme title, teaching hours, completion date and unique verification reference.
• Verified identity, full attendance and separate written and applied passes are required; all mandatory critical criteria must be met.
• Proposed issue target: within 30 calendar days of approved results and settlement of applicable fees.
• Records assessed introductory applied machine learning skills; it does not confer data scientist certification or authorise deployment into production systems.
• Attendance confirmation may record completed hours for unsuccessful learners; it does not certify successful completion.

Programme Profile

Number of daysNumber of daysNumber of days

555

Total teaching hoursTotal teaching hoursTotal teaching hours

35 hours; breaks, pre course work and formal final assessments excluded.35 hours; breaks, pre course work and formal final assessments excluded.35 hours; breaks, pre course work and formal final assessments excluded.

Delegates allowedDelegates allowedDelegates allowed

Maximum 16 delegates.Maximum 16 delegates.Maximum 16 delegates.

Training prerequisitesTraining prerequisitesTraining prerequisites

Basic Python programming, dataframe manipulation and introductory statistics are required. Familiarity with functions, arrays, averages and variability is expected; this is not an introductory programming course.Basic Python programming, dataframe manipulation and introductory statistics are required. Familiarity with functions, arrays, averages and variability is expected; this is not an introductory programming course.Basic Python programming, dataframe manipulation and introductory statistics are required. Familiarity with functions, arrays, averages and variability is expected; this is not an introductory programming course.

Pre course assignment and assessmentPre course assignment and assessmentPre course assignment and assessment

Complete a 45 minute Python and statistics diagnostic and verify the supplied notebook environment before attendance.Complete a 45 minute Python and statistics diagnostic and verify the supplied notebook environment before attendance.Complete a 45 minute Python and statistics diagnostic and verify the supplied notebook environment before attendance.

End of course assessmentEnd of course assessmentEnd of course assessment

100 question written exam (150 minutes) and individual applied task (90 minutes).100 question written exam (150 minutes) and individual applied task (90 minutes).100 question written exam (150 minutes) and individual applied task (90 minutes).

Passing criteriaPassing criteriaPassing criteria

75% written and 70% applied, passed separately; full attendance. All mandatory criteria must pass.75% written and 70% applied, passed separately; full attendance. All mandatory criteria must pass.75% written and 70% applied, passed separately; full attendance. All mandatory criteria must pass.