SOA ATPA – Advanced Topics in Predictive Analytics

Develop Advanced Expertise in Predictive Modelling, Machine Learning, Data Analytics and Professional Communication

Advanced Topics in Predictive Analytics (ATPA) is an e-learning and assessment programme conducted by the Society of Actuaries (SOA), USA, forming part of the educational pathway towards the Associate of the Society of Actuaries (ASA) designation.

The programme focuses on applying advanced data science, statistical learning and predictive modelling techniques to solve complex business problems using real-world datasets.

ATPA builds upon the statistical and analytical knowledge developed in Exam SRM – Statistics for Risk Modeling and Exam PA – Predictive Analytics.

The programme combines data preparation, advanced predictive models, neural networks, Bayesian techniques, model validation, ethical considerations, explainable artificial intelligence and professional communication, developing analytical and technical skills for actuarial practice.

Unlike conventional actuarial examinations, ATPA is assessed through a take-home practical assessment requiring candidates to analyse data, develop predictive models and submit a professional report.

Course Structure & Syllabus

ATPA combines advanced predictive analytics concepts with practical applications in statistical modelling, machine learning, risk assessment and business decision-making.

Key Topics Covered:

1. Ethical Foundations & Responsible Analytics:
   Ethical use of data, professional standards, regulatory considerations, fairness, discrimination, responsible modelling and ethical decision-making.
2. Data Sources, Pipelines & Database Management:
   Data sources, databases, data warehouses, data lakes, relational databases, data structures and Extract, Transform and Load (ETL) processes.
3. Data Preparation & Feature Engineering:
   Data cleaning, missing-value treatment, outlier identification, data transformations, merging datasets, variable creation, data quality and detection of potential biases.
4. Advanced Statistical & Predictive Models:
   Generalised additive models, linear mixed models, neural networks, model fitting, parameter estimation and predictive modelling applications.
5. Bayesian Modelling & Ensemble Techniques:
   Bayesian predictive techniques, model combination, stacking, blending, comparative model performance and selection of appropriate modelling methods.
6. Model Validation, Selection & Risk Assessment:
   Model accuracy, cross-validation, overfitting, model stability, computational efficiency, fairness, model bias and evaluation of business suitability.
7. Model Explainability & Professional Communication:
   Variable importance, partial dependence plots, individual conditional expectation plots, Shapley values, lift charts, gain charts, model documentation and professional reporting.

The ATPA syllabus focuses on advanced predictive modelling, ethical analytics, data engineering, machine learning, model explainability and effective communication of analytical findings.

Syllabus Topics & Weightages

The SOA ATPA syllabus comprises four principal areas:

These weightages are indicative ranges from the SOA ATPA syllabus rather than fixed percentages. Individual assessment tasks may combine several learning objectives, and the distribution may vary between assessment versions.

Module Structure & Assessment

The ATPA programme includes four e-learning modules that support candidates in developing advanced predictive analytics knowledge and preparing for the comprehensive assessment.

Assessment Structure

ATPA is assessed through a take-home practical assignment involving a business problem, a dataset, predictive modelling and a professional written report.

Important Assessment Guidelines

• Take-Home Assessment:
  Candidates receive a realistic business problem and dataset requiring statistical analysis, predictive modelling and professional recommendations.
• Assessment Time Limit:
  Candidates must complete and submit their assessment within 96 hours of downloading it or before the published assessment submission deadline, whichever occurs first.
• Statistical Modelling:
  Candidates may apply advanced predictive analytics techniques, including statistical models, neural networks, Bayesian methods and ensemble modelling approaches.
• Programming Software:
  Candidates may use R, Python or other suitable statistical software. They are responsible for arranging access to the software and computational resources required for their work.
• Professional Reporting:
  The required submission is a written report presenting the analytical approach, relevant findings, model interpretations and appropriate recommendations.
• External Resources:
  Candidates may consult permitted reference materials and external resources in accordance with SOA’s assessment instructions and academic integrity requirements.
• Model Explainability:
  Candidates should demonstrate an understanding of model assumptions, limitations, fairness, interpretability and the business implications of modelling results.
• Assessment Integrity:
  All submitted work must comply with SOA’s rules regarding independent work, collaboration, confidentiality, plagiarism and assessment security.
• Assessment Registration:
  Candidates must register with the SOA for the applicable ATPA assessment version and comply with the published registration and submission deadlines.

Assessment Schedule

The SOA publishes separate registration periods, assessment availability dates, submission deadlines and result release schedules for ATPA.

Dates and deadlines are subject to SOA updates. The October–December 2026 assessment registration closes on 30 November 2026. Submission deadlines are based on US Central Time. Candidates must also meet the 96-hour assessment clock.

Eligibility Criteria

ATPA is suitable for actuarial students developing advanced statistical learning, predictive modelling and business analytics skills.

• A strong understanding of SOA Exam SRM – Statistics for Risk Modeling and Exam PA – Predictive Analytics is recommended.
• Knowledge of Exam P – Probability and VEE Mathematical Statistics concepts is assumed.
• Familiarity with regression analysis, statistical learning, model validation, data exploration and predictive analytics is beneficial.
• Candidates should be comfortable using R, Python or other statistical software for data analysis and predictive modelling.
• Students from Actuarial Science, Statistics, Mathematics, Data Science, Finance, Economics and other quantitative backgrounds may prepare for the assessment.
• Candidates must meet the SOA’s applicable registration, identification, assessment and submission requirements.

Learning Objectives

Studying ATPA develops the ability to:

• Apply ethical principles to data science and predictive modelling.
• Evaluate data sources, data quality and analytical requirements.
• Understand databases, data warehouses and data pipelines.
• Perform data cleaning, transformation and feature preparation.
• Develop advanced statistical and machine learning models.
• Apply neural networks, Bayesian methods and ensemble techniques.
• Evaluate model performance, fairness, stability and limitations.
• Use model explainability techniques to interpret predictions.
• Apply professional judgement to predictive analytics decisions.
• Communicate statistical findings through structured professional reports.

Practical Applications of ATPA

The knowledge gained through ATPA is applicable across several actuarial, financial and analytical disciplines:

1. General Insurance:
   Developing predictive pricing models, identifying claim risk factors, analysing policyholder behaviour and forecasting insurance losses.
2. Life & Health Insurance:
   Modelling mortality and morbidity risks, evaluating customer behaviour and applying predictive analytics to insurance experience.
3. Banking & Financial Risk Management:
   Developing credit risk models, analysing financial patterns, evaluating customer risks and supporting quantitative decision-making.
4. Data Science & Machine Learning:
   Applying neural networks, statistical models, ensemble learning, explainable AI and advanced data preparation techniques to complex datasets.
5. Enterprise Risk Management & Consulting:
   Identifying emerging risks, developing predictive models, assessing model uncertainty and communicating analytical recommendations to business stakeholders.

ATPA develops advanced predictive analytics and professional communication skills, connecting statistical learning with practical business applications and supporting progression towards the Associate of the Society of Actuaries (ASA) designation.

Technical Requirements – AEI Learning Application

To access online classes, recorded lectures and digital study resources, students must ensure their devices meet the following minimum technical requirements of the Actuators Educational Institute (AEI) Learning Application.

Important Information

ATPA is an independently assessed SOA e-learning programme. AEI preparation supports conceptual understanding, data analysis, statistical modelling, programming skills, practice scenarios and professional communication.

Candidates must register separately with the Society of Actuaries for the official ATPA Assessment and complete their submissions independently in accordance with SOA assessment-integrity requirements.

The SOA may revise assessment procedures, learning objectives, registration fees and submission schedules. Candidates should verify the official SOA guidance applicable to their assessment version.