اردو

Home / Practice / SDS Problem identification and data readiness / Worksheet

Practise online

SDS Problem identification and data readiness

Written by Saad Rahman (ACCA)

ACCA · New at Strategic Professional level · 7 questions

Name: ____________________   Date: ____________

  1. In a data science project, the first stage of CRISP-ML is:

    • A Monitoring
    • B Business and data understanding
    • C Deployment
    • D Modelling

  2. Which statement describes a well-scoped data science problem?

    • A It starts with the choice of algorithm
    • B It has no stakeholders
    • C It states a specific decision, a measure of success and the data available
    • D It says 'use AI to improve results'

  3. A data quality audit would check completeness, accuracy, consistency and:

    • A The server brand
    • B The number of users
    • C Timeliness and validity
    • D The colour of the charts

  4. A model is built on five years of customer data collected only from branch customers, and will be used for online customers. The main concern is:

    • A The model is too fast
    • B The users are too many
    • C The data may not be representative
    • D The data is too clean

  5. A deployed model's accuracy falls over time because customer behaviour has changed. This is called:

    • A Underfitting
    • B Data leakage
    • C Model drift
    • D Overfitting

  6. Early stakeholder engagement in a data science project is important mainly because it:

    • A Reduces the need for data
    • B Allows the team to avoid documentation
    • C Ensures that the work addresses a real need and increases the chance the result is used
    • D Avoids all risk

  7. Which of these is a data preparation task?

    • A Handling missing values and encoding categories
    • B Presenting results to the board
    • C Monitoring drift
    • D Agreeing the problem

Answer key

1. B, 2. C, 3. C, 4. C, 5. C, 6. C, 7. A

  1. The project starts by defining the problem and checking whether suitable data exists.

  2. A well-scoped problem links to value and can be measured.

  3. These are standard dimensions of data quality, and uniqueness is also commonly tested.

  4. Models perform poorly when training data differs from the population to which it is applied.

  5. Models need monitoring and maintenance after deployment.

  6. Poor engagement leads to models that are never adopted.

  7. Preparation makes the data suitable for modelling.