Data Cleaning and Preprocessing

Active learning works especially well for data cleaning and preprocessing because students need to experience the real consequences of messy data. Handling errors themselves builds an intuitive grasp of why each method matters, which lectures alone cannot achieve.

• Class 11
• Computer Science
• 4 activities
• 25 min–45 min

Learning Objectives

1. Identify types of data errors, including missing values, outliers, and inconsistencies, within a given dataset.
2. Compare and contrast at least two methods for handling missing data, such as deletion and mean imputation.
3. Critique a sample dataset to pinpoint potential data quality issues and propose specific cleaning strategies.
4. Demonstrate the application of a chosen data cleaning technique to rectify errors in a small dataset using a spreadsheet or basic programming tool.

Collaborative Problem-Solving

⏱ 30 min·Pairs

Pair Work: Dataset Audit

Provide pairs with a sample dataset containing missing values and inconsistencies. They list errors, choose handling methods, and apply fixes using spreadsheets. Pairs then swap datasets to verify each other's work.

Prepare & details

Explain why data cleaning is a critical step before data analysis.

Facilitation Tip: During Pair Work: Dataset Audit, provide colour-coded printouts so pairs can physically circle errors before deciding how to fix them, making the process visual and collaborative.

Setup: Flexible seating that allows clusters of 5-6 students; desks can be grouped in rows of three facing each other if fixed furniture limits rearrangement. Wall or board space for displaying group norm charts and the session agenda is helpful.

Materials: Printed problem brief cards (one per group), Role cards: Facilitator, Questioner, Recorder, Devil's Advocate, Communicator, Group norm chart (printable poster format), Individual reflection sheet and exit ticket, Timer visible to the class (board countdown or projected timer)

• Apply
• Analyze
• Evaluate
• Create
• Relationship Skills
• Decision-Making
• Self-Management

Collaborative Problem-Solving

⏱ 45 min·Small Groups

Small Groups: Outlier Detection Challenge

Distribute datasets with outliers to small groups. Groups plot data, use IQR to identify outliers, and decide on removal or adjustment. They present findings and rationale to the class.

Prepare & details

Differentiate between various techniques for handling missing data.

Facilitation Tip: During Small Groups: Outlier Detection Challenge, give each group a ruler to measure distances on printed box plots so they connect statistical positions to visual outliers.

Setup: Flexible seating that allows clusters of 5-6 students; desks can be grouped in rows of three facing each other if fixed furniture limits rearrangement. Wall or board space for displaying group norm charts and the session agenda is helpful.

Materials: Printed problem brief cards (one per group), Role cards: Facilitator, Questioner, Recorder, Devil's Advocate, Communicator, Group norm chart (printable poster format), Individual reflection sheet and exit ticket, Timer visible to the class (board countdown or projected timer)

• Apply
• Analyze
• Evaluate
• Create
• Relationship Skills
• Decision-Making
• Self-Management

Collaborative Problem-Solving

⏱ 35 min·Whole Class

Whole Class: Cleaning Simulation

Project a large messy dataset. Class votes on issues via hand signals, then brainstorms strategies collectively. Implement top ideas live and discuss impact on summary statistics.

Prepare & details

Critique a dataset for potential errors and propose cleaning strategies.

Facilitation Tip: During Whole Class: Cleaning Simulation, assign roles like 'data owner' and 'cleaner' so students hear each other articulate trade-offs between deletion and imputation.

Setup: Flexible seating that allows clusters of 5-6 students; desks can be grouped in rows of three facing each other if fixed furniture limits rearrangement. Wall or board space for displaying group norm charts and the session agenda is helpful.

Materials: Printed problem brief cards (one per group), Role cards: Facilitator, Questioner, Recorder, Devil's Advocate, Communicator, Group norm chart (printable poster format), Individual reflection sheet and exit ticket, Timer visible to the class (board countdown or projected timer)

• Apply
• Analyze
• Evaluate
• Create
• Relationship Skills
• Decision-Making
• Self-Management

Collaborative Problem-Solving

⏱ 25 min·Individual

Individual: Personal Data Clean-Up

Students collect class survey data individually, clean it for missing entries and outliers, then compute basic statistics. Share cleaned versions in a class repository for comparison.

Prepare & details

Explain why data cleaning is a critical step before data analysis.

Facilitation Tip: During Individual: Personal Data Clean-Up, set a strict 15-minute timer so students feel the pressure of real-world constraints and prioritise fixes accordingly.

Setup: Flexible seating that allows clusters of 5-6 students; desks can be grouped in rows of three facing each other if fixed furniture limits rearrangement. Wall or board space for displaying group norm charts and the session agenda is helpful.

Materials: Printed problem brief cards (one per group), Role cards: Facilitator, Questioner, Recorder, Devil's Advocate, Communicator, Group norm chart (printable poster format), Individual reflection sheet and exit ticket, Timer visible to the class (board countdown or projected timer)

• Apply
• Analyze
• Evaluate
• Create
• Relationship Skills
• Decision-Making
• Self-Management

Teaching This Topic

Teachers should treat this topic as skill-building rather than theory. Avoid long lectures on methods; instead, let students fail with dirty data first, then guide them to discover corrections. Research shows that students retain cleaning techniques when they first experience the pain of unclean data themselves, so structure activities where errors have visible consequences.

What to Expect

Successful learning looks like students confidently identifying errors in raw data and justifying their chosen cleaning method. You will see them comparing 'before and after' datasets to prove that preprocessing improves data quality.

Watch Out for These Misconceptions

Common Misconception

During Pair Work: Dataset Audit, watch for students who immediately delete rows with missing values without discussing bias.

What to Teach Instead

Ask pairs to calculate the percentage of missing data and compare datasets before and after deletion; this will reveal how deletion shrinks the dataset and may skew results.

Common Misconception

During Small Groups: Outlier Detection Challenge, watch for students who remove all outliers without checking if they are valid extremes.

What to Teach Instead

Have groups plot the data before and after outlier removal and present whether the trend line changes significantly; this forces them to justify removal decisions.

Common Misconception

During Individual: Personal Data Clean-Up, watch for students who skip cleaning small datasets like class surveys, assuming errors are trivial.

What to Teach Instead

Require them to run a simple summary statistic (mean, median) before and after cleaning to demonstrate how even small errors shift results.

Assessment Ideas

Quick Check

After Pair Work: Dataset Audit, give pairs a fresh table with deliberate errors and ask them to identify two errors and suggest one correction for each.

Discussion Prompt

During Whole Class: Cleaning Simulation, pose: 'If you were building a recommendation system for an e-commerce website, what kinds of data cleaning challenges might you face with user purchase history?' Ask students to share their proposed solutions in a class discussion.

Exit Ticket

During Individual: Personal Data Clean-Up, hand out scenario cards (e.g., cleaning data for a weather forecast model) and ask students to write one data quality issue, the technique they would use, and why it fits the scenario.

Extensions & Scaffolding

• Challenge: Ask early finishers to create a new dataset from scratch with deliberate errors and swap with another student for cleaning practice.
• Scaffolding: Provide a checklist of common errors (missing values, outliers, inconsistent formats) that students can tick off as they clean their personal dataset.
• Deeper: Invite students to research a real-world case where poor data cleaning led to public consequences, then present findings to the class.

Key Vocabulary

Missing Values	Data points that are absent or not recorded for a particular observation. These can occur due to errors in data entry or collection.
Outliers	Data points that significantly differ from other observations in a dataset. They can be due to measurement errors or represent genuine extreme values.
Inconsistencies	Discrepancies or contradictions within a dataset, such as different formats for the same information (e.g., 'New Delhi' vs. 'Delhi, India') or illogical entries.
Data Imputation	The process of replacing missing data values with substituted values. Common methods include using the mean, median, or mode of the available data.
Data Normalization/Standardization	Techniques used to rescale data to a common range or distribution, often to prepare it for analysis or machine learning algorithms. This helps in handling inconsistencies in units or scales.