Capstone Project Guide

Comprehensive guidelines for the Data Science & AI Capstone Project based on real organizational problems, data ethics, and deployment to the public cloud.

Module Author: Mahendar Dwi Payana
Syllabus Reference: Thematic Academy DTS Kominfo RI - STEI ITB & SKKNI No. 299 of 2020


1. Main Objective

The Capstone Project is the proof of integration for all 16 learning outcomes mastered during the 8-week program.

Participants are challenged to apply the End-to-End Data Science Lifecycle to solve real problems faced by institutions, universities, government agencies, or industry.


2. Team & Governance Requirements

  • Group Formation: Each group consists of 3 members (or 4 members for one group if the participant count is odd).
  • Suggested Team Roles:
    • Data / ML Engineer: Responsible for data acquisition, pre-processing, scaling pipeline, and API deployment.
    • Data Scientist / Modeler: Responsible for problem formulation, EDA, algorithm experiments, and model tuning validation.
    • Full-Stack / Product Lead: Responsible for web interface design, visualization integration, and the business evaluation report.

3. Dataset Selection Ethics (Critical)

The committee and curriculum enforce strict rules on data integrity:

  1. Own Organizational Real Data: Internal data from universities, research labs, employers, or business partners.
  2. National Open Data: Official data from Indonesian government portals, such as:
  3. Ethics & Anonymization: Any data containing confidential personal identity (Personally Identifiable Information - PII) must be masked or encrypted before use.

B. Strict Prohibitions (Automatic Disqualification)

[!CAUTION] It is strictly forbidden to use clean, popular toy datasets commonly found in online tutorials, such as:

  • General Kaggle Datasets (Titanic, Iris, House Prices, generic Credit Card Fraud).
  • The UCI Machine Learning Repository (archive.ics.uci.edu).
  • Old academic benchmark datasets (plain MNIST, CIFAR-10, Boston Housing).

4. 8-Week Milestone Schedule

WeekSession DatesLearning FocusDeliverableDeadline
1Jul 31 - Aug 1Team FormationGroup Member & Role Assignment FormAug 6
2Aug 7 - 8Data Science MethodologyProblem Description Draft & Business ObjectiveAug 13
3Aug 14 - 15Business & Data UnderstandingRevised Problem Statement & Initial Data StatisticsAug 20
4Aug 21 - 22Visualization & CleaningEDA Report & Data Cleaning LogAug 27
5Aug 28 - 29Construction & Feature EngineeringProcessed Dataset (Scaling, Encoding, Feature Selection)Sep 3
6Sep 4 - 5Classification & Regression ModelsSupervised Training Scripts & Experiment ScoreboardSep 10
7Sep 11 - 12Clustering & ANN ModelsSegmentation / Neural Network Model ImplementationSep 17
8Sep 18 - 19Evaluation & DeploymentLive Cloud Web App Link + Technical ReportSep 26

5. Final Assessment Criteria (50% of Passing Grade)

  1. Business Relevance (15%): Sharpness in mapping a real business problem into a measurable machine learning formulation.
  2. Data Preparation & EDA Rigor (25%): Data cleanliness, rational imputation, creative feature engineering, and freedom from data leakage.
  3. Modeling & Validation Rigor (25%): Appropriate evaluation metrics, Stratified Cross-Validation scenarios, and fair algorithm comparison.
  4. Operational Web App Quality (25%): Ease of use for non-technical users, stable prediction inference, and cloud response time.
  5. Documentation & Final Report (10%): Clarity of the GitHub repository structure, README, and scientific conclusions.

On this page